Files
lab/migration/vrrp-wan-reconcile

175 lines
7.9 KiB
Plaintext
Raw Normal View History

vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
#!/bin/sh
# Make the WAN match VRRP mastership. Idempotent; safe to run every 30s and on
# every VRRP transition.
#
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
# Two WANs, two different control planes, for a reason:
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
#
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
# bond0.53 (10 gig, DHCP) -- CONFIG plane. Its lease is bound to a cloned MAC
# (f0:9f:c2:12:9b:4f, the old USG's), and only VyOS
# config can move a MAC. One commit per failover.
# pppoe0 (Vodafone) -- SYSTEMD plane. Gated by a drop-in on
# ppp@pppoe0; see migration/ppp-vrrp-gate.conf.
# No commit, no config lock, no `save`.
#
# PPPoE used to be on the config plane too, via `set interfaces pppoe pppoe0
# disable`. That could not work: `disable` unlinks /etc/ppp/peers/pppoe0, which
# is pppd's options file, so the promotion path deleted the very thing it needed
# and left the unit restart-looping (observed: 47 restarts, zero sessions at the
# access concentrator).
#
# Why a reconciler and not just transition scripts: VyOS delivers
# `transition-script` through keepalived-fifo.py, and on 2026-09-02 that helper
# logged NOTHING for a promotion while Keepalived_vrrp logged all six instances
# entering MASTER. A router held every VIP with no WAN -- the outage, recreated
# by the mechanism meant to prevent it. Scripts give speed; the timer gives
# correctness.
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
#
# vrrp-wan-reconcile reconcile once
# vrrp-wan-reconcile --status what it thinks, changing nothing
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
CONF=/config/vrrp-wan.conf
[ -r "$CONF" ] && . "$CONF"
VIP="${VRRP_WAN_VIP:-192.168.1.1}"
WAN_VIF="${WAN_VIF:-53}"
FLAP_MAX="${FLAP_MAX:-6}"
FLAP_WINDOW="${FLAP_WINDOW:-600}"
FLAP_HOLDOFF="${FLAP_HOLDOFF:-900}"
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
STATE=/run/vrrp-wan
LOCK=/run/vrrp-wan.lock
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
APPLY=/config/vrrp-wan-apply
DROPIN=/etc/systemd/system/ppp@pppoe0.service.d/10-vrrp-wan-gate.conf
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
cfg() { /opt/vyatta/bin/vyatta-op-cmd-wrapper show configuration commands 2>/dev/null; }
holds_vip() { ip -4 -o addr show 2>/dev/null | grep -q " ${VIP}/"; }
wan_up() { ip -4 addr show "bond0.${WAN_VIF}" 2>/dev/null | grep -q 'inet '; }
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
ppp_up() { ip -4 addr show pppoe0 2>/dev/null | grep -q 'inet '; }
ppp_active() { systemctl is-active --quiet ppp@pppoe0 2>/dev/null; }
lease_age() { s=$(stat -c %Y "$STATE/may-dial" 2>/dev/null) || return 1
echo $(( $(date +%s) - s )); }
# NOTE: there is deliberately no ppp_disabled(). pppoe0 is now ENABLED in config
# on both routers, so such a test would be permanently false and the backup
# early-exit below would never fire -- entering config mode every 30s for ever,
# committing nothing. That exact shape was already live on the sim secondary,
# whose /tmp/vrrp-wan-commit.log read "No configuration changes to commit" while
# the script reported success.
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
wan_disabled(){ cfg | grep -q "vif ${WAN_VIF} disable"; }
if [ "${1:-}" = "--status" ]; then
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
printf 'vip=%s holds_vip=%s wan_disabled=%s wan_up=%s ppp_up=%s ppp_active=%s may_dial=%s lease_age=%s dropin=%s role=%s\n' \
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
"$VIP" "$(holds_vip && echo yes || echo no)" \
"$(wan_disabled && echo yes || echo no)" \
"$(wan_up && echo yes || echo no)" \
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
"$(ppp_up && echo yes || echo no)" \
"$(ppp_active && echo yes || echo no)" \
"$([ -f "$STATE/may-dial" ] && echo yes || echo no)" \
"$(lease_age 2>/dev/null || echo -)" \
"$([ -f "$DROPIN" ] && echo yes || echo MISSING)" \
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
"$(cat "$STATE/role" 2>/dev/null || echo unset)"
exit 0
fi
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
# One writer. The lock fd MUST be closed for children (`9>&-` on every call):
# entering VyOS config mode spawns a long-lived unionfs-fuse for the session
# which INHERITS the descriptor and never releases it, so from the first commit
# onward every later run lost the flock and exited 0 having done nothing. That
# is how a demoted router kept the WAN.
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
exec 9>"$LOCK"
flock -n 9 || exit 0
mkdir -p "$STATE"
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
# Reap config sessions whose owning process is gone. VyOS leaves a unionfs mount
# per `configure`, one of them holds the commit lock, and after that EVERY commit
# fails -- including the manual one you try in order to fix it.
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
for d in /opt/vyatta/config/tmp/new_config_*; do
[ -d "$d" ] || continue
pid=${d##*_}
kill -0 "$pid" 2>/dev/null && continue
umount -l "$d" 2>/dev/null
rm -rf "$d" 2>/dev/null
done
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
# `show configuration commands` has been observed returning EMPTY transiently
# under commit-lock contention. Every grep against it then reads false, which on
# the master path looks like "the WAN is disabled" and triggers a pointless
# commit -- one such spurious "releasing" was logged on a box where both WANs
# were already in the right state. A real config is ~500 lines; refuse to act on
# a suspiciously short one.
if [ "$(cfg | wc -l)" -lt 50 ]; then
logger -t vrrp-wan "config read returned <50 lines; skipping this tick"
exit 0
fi
# --- PPPoE: the systemd plane ---------------------------------------------
ppp_dial() {
# Refuse to bless a box whose gate is missing. /etc is per-image, so a VyOS
# upgrade silently drops the drop-in -- and without it BOTH routers dial on
# the next commit that touches the pppoe subtree. Failing closed turns a
# silent loss of protection into "PPPoE never comes up", the safe direction.
if [ ! -f "$DROPIN" ]; then
logger -t vrrp-wan "REFUSING to dial: gate drop-in $DROPIN is missing (VyOS upgrade?)"
return 1
fi
now=$(date +%s)
if [ -f "$STATE/holdoff" ] && [ "$now" -lt "$(cat "$STATE/holdoff" 2>/dev/null || echo 0)" ]; then
return 1
fi
touch "$STATE/may-dial" # renew the lease every tick
ppp_active && return 0
# Trim the dial log to the window, then decide.
if [ -f "$STATE/dials" ]; then
awk -v c="$((now - FLAP_WINDOW))" '$1 > c' "$STATE/dials" > "$STATE/dials.new" 2>/dev/null
mv "$STATE/dials.new" "$STATE/dials" 2>/dev/null
fi
if [ "$(wc -l < "$STATE/dials" 2>/dev/null || echo 0)" -ge "$FLAP_MAX" ]; then
echo $((now + FLAP_HOLDOFF)) > "$STATE/holdoff"
logger -t vrrp-wan "DIAL FLAP: >=${FLAP_MAX} attempts in ${FLAP_WINDOW}s -- holding off ${FLAP_HOLDOFF}s"
return 1
fi
echo "$now" >> "$STATE/dials"
logger -t vrrp-wan "MASTER: dialling pppoe0"
systemctl reset-failed ppp@pppoe0 2>/dev/null
# `systemctl start` exits 0 even when a Condition blocks the start, so its
# return code proves nothing. is-active is the only honest answer.
systemctl start ppp@pppoe0 2>/dev/null
}
ppp_release() {
# Order matters: revoke the lease FIRST, then stop. The file's absence blocks
# any NEW start (including one a concurrent VyOS commit would trigger); the
# stop kills the process that already exists. Stopping first leaves a window
# in which a commit re-dials a box that is being demoted.
rm -f "$STATE/may-dial"
ppp_active || return 0
logger -t vrrp-wan "not MASTER: hanging up pppoe0"
systemctl stop ppp@pppoe0 2>/dev/null
}
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
# --- decide ----------------------------------------------------------------
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
if holds_vip; then
echo master > "$STATE/role"
[ -f "$STATE/since" ] || date +%s > "$STATE/since"
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
ppp_dial
wan_disabled || exit 0
logger -t vrrp-wan "MASTER with bond0.${WAN_VIF} disabled -> enabling"
t0=$(date +%s)
vyos: a failover you can actually trigger, and four bugs found triggering it A planned failover had no reliable lever. VyOS offers only `restart vrrp`, and neither that nor `systemctl restart keepalived` is dependable: with advert_int 1 the peer declares the master dead after ~3.6s and a restart usually finishes inside that window. Measured -- the same command moved mastership on one run and not on the next three. A fail-back step you cannot trigger on purpose is not a procedure, and the recovery card depended on one. The lever is now `touch /run/vrrp-wan/force-fault`: the health check fails, the sync group sheds every VIP, and the peer takes over. It exercises the same path a real WAN loss takes rather than a special case, and it lives in /run so a reboot cannot leave a router permanently ineligible. Proven end to end in labsim: lever -> mastership moves -> WAN follows -> the old master releases -> a LAN VM has internet -> the faulted router returns to BACKUP and is eligible again. Getting there exposed four real bugs, two of which would have broken a GENUINE failover, not just the drill: - The grace stamp was written only by the 30s reconciler, so a freshly promoted master reached the 5s health check with no stamp, scored grace = 0, failed instantly and went FAULT. With the peer already faulted that left BOTH routers in FAULT and the LAN with no gateway at all -- worse than the outage the check exists to prevent. The check now stamps on promotion. - And it inherited STALE stamps from an earlier mastership, failing ~5s after passing. The stamp is now cleared on the way down, by the health check itself, not only by the reconciler. - The lock fd leaked into VyOS's config session: `exec 9>` is inherited by the long-lived unionfs-fuse the session spawns, which never closes it. From the first config change on, every later reconciler run lost the flock and exited 0 having done nothing -- healthy-looking journal, silently stopped reconciling. That is how a demoted router kept the WAN. Children now get 9>&-. - vrrp-wan-apply touched pppoe0 unconditionally. On a box where pppoe0 has no source-interface VyOS rejects the whole commit ("Physical source-interface required"), taking the bond0.53 change down with it -- and the script still returned 0, so the reconciler logged a release that never happened. pppoe0 is now guarded on existence and the commit's verdict is propagated. Still NOT applied to production. The pair is single-homed on WAN until it is. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 23:27:17 +01:00
"$APPLY" enable 9>&-
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
logger -t vrrp-wan "bond0.${WAN_VIF} enable commit took $(( $(date +%s) - t0 ))s"
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
else
echo backup > "$STATE/role"
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
rm -f "$STATE/since" "$STATE/holdoff"
ppp_release
wan_disabled && exit 0
logger -t vrrp-wan "not MASTER but bond0.${WAN_VIF} enabled -> releasing"
t0=$(date +%s)
vyos: a failover you can actually trigger, and four bugs found triggering it A planned failover had no reliable lever. VyOS offers only `restart vrrp`, and neither that nor `systemctl restart keepalived` is dependable: with advert_int 1 the peer declares the master dead after ~3.6s and a restart usually finishes inside that window. Measured -- the same command moved mastership on one run and not on the next three. A fail-back step you cannot trigger on purpose is not a procedure, and the recovery card depended on one. The lever is now `touch /run/vrrp-wan/force-fault`: the health check fails, the sync group sheds every VIP, and the peer takes over. It exercises the same path a real WAN loss takes rather than a special case, and it lives in /run so a reboot cannot leave a router permanently ineligible. Proven end to end in labsim: lever -> mastership moves -> WAN follows -> the old master releases -> a LAN VM has internet -> the faulted router returns to BACKUP and is eligible again. Getting there exposed four real bugs, two of which would have broken a GENUINE failover, not just the drill: - The grace stamp was written only by the 30s reconciler, so a freshly promoted master reached the 5s health check with no stamp, scored grace = 0, failed instantly and went FAULT. With the peer already faulted that left BOTH routers in FAULT and the LAN with no gateway at all -- worse than the outage the check exists to prevent. The check now stamps on promotion. - And it inherited STALE stamps from an earlier mastership, failing ~5s after passing. The stamp is now cleared on the way down, by the health check itself, not only by the reconciler. - The lock fd leaked into VyOS's config session: `exec 9>` is inherited by the long-lived unionfs-fuse the session spawns, which never closes it. From the first config change on, every later reconciler run lost the flock and exited 0 having done nothing -- healthy-looking journal, silently stopped reconciling. That is how a demoted router kept the WAN. Children now get 9>&-. - vrrp-wan-apply touched pppoe0 unconditionally. On a box where pppoe0 has no source-interface VyOS rejects the whole commit ("Physical source-interface required"), taking the bond0.53 change down with it -- and the script still returned 0, so the reconciler logged a release that never happened. pppoe0 is now guarded on existence and the commit's verdict is propagated. Still NOT applied to production. The pair is single-homed on WAN until it is. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 23:27:17 +01:00
"$APPLY" disable 9>&-
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
logger -t vrrp-wan "bond0.${WAN_VIF} disable commit took $(( $(date +%s) - t0 ))s"
vyos: WAN follows VRRP mastership, rehearsed in labsim The ISP is consumer with one static IP, so "both routers hold WAN" is not available: the 10 gig lease is anchored to a cloned MAC and the PPPoE line to a single credential. The WAN therefore has to move with mastership. Proven end to end in the sim: the secondary was promoted, took the WAN, got a lease, installed a default route, and a LAN VM reached the internet through it (3/3, 9ms). Demoting released it -- link down, no address. Rebooting the master converged correctly too: it came back BACKUP with the WAN disabled while the peer kept it. The rehearsal earned its keep four times over, and none of these were visible from reading the docs: - `transition-script` alone is NOT safe to hang internet on. VyOS delivers it through keepalived-fifo.py, and on one promotion that helper logged NOTHING while Keepalived_vrrp logged all six instances entering MASTER and the built-in notify_master for conntrack-sync ran normally. The result was a router holding every VIP with no WAN -- the 2026-09-02 outage, recreated by the mechanism meant to prevent it. Hence vrrp-wan-reconcile on a 30s timer: the scripts give speed, the timer gives correctness. - The health check must ask REALITY, not a marker. Keying "am I master" on a /run file written by the transition script meant that when the script did not run, the router believed it was backup, passed the check, and kept the VIPs it could not serve. It now asks whether the VIP is actually on the box. - The old address-based check DEADLOCKED this design: may-I-be-master required already having WAN, and only the master gets WAN. That is why vyos002 sat in FAULT for ever -- the safety check had silently removed the redundancy it existed to protect. - script-template must be the FIRST thing a script does. Sourced after an if, an exec and a mkdir it terminated the script inside the source, rc=0, no output: the reconciler reported success having done nothing. Hence the split into vrrp-wan-apply, matching the shape /config/vyos-known-good already uses. Two hazards found and handled rather than discovered in production: - A `configure` session whose process dies leaks a unionfs mount under /opt/vyatta/config/tmp, and one of those holds the commit lock -- after which every commit fails, including the manual one you try to fix it with. A 30s job that can leak one per failure wedges the box on its own, so the reconciler reaps dead sessions before it starts. It cleared 8 on the sim. - Any `save` while a box is master persists the enabled WAN into config.boot, so a reboot would claim the shared MAC regardless of VRRP state. Observed: an ordinary console-apply did exactly this. config.boot must keep `disable` on BOTH routers; the model asserts it and vyos:verify reports it as drift. NOT yet applied to production, and it should not be until the remaining item is settled: a clean, deliberately-triggered failover has been seen via reboot, but `restart vrrp` twice failed to move mastership at all, so the trigger for a planned failover is still unproven. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-02 18:02:58 +01:00
fi
vyos: move PPPoE off the config plane onto a gated systemd unit PPPoE HA could not work as written, and the reason is structural rather than a bug: `set interfaces pppoe pppoe0 disable` and `delete` are handled identically by interfaces_pppoe.py -- both UNLINK /etc/ppp/peers/pppoe0. That path is pppd's own options file, so the resting state destroyed exactly what the promotion path needed, and `ppp@pppoe0` restart-looped against it (observed: 47 restarts, zero sessions at the access concentrator). It also made op-mode `connect interface pppoe0` unusable, and put every failover behind a priority-322 commit where one unrelated invalid node fails the whole thing -- which has already taken the 10 gig down once. pppoe0 is now configured identically and ENABLED on both routers, so the peers file always exists, and dialling is gated by a drop-in on the unit: ConditionPathExists=/run/vrrp-wan/may-dial ConditionPathExists=/etc/ppp/peers/pppoe0 /run is tmpfs, so the gate is shut at boot and neither box can dial before VRRP has decided. That matters more than it first appears: with the node enabled, interfaces_pppoe.py restarts ppp on EVERY commit touching the pppoe subtree when the daemon is not running -- so the backup actively tries to dial whenever anything commits. The gate is the only thing making that a no-op, which is why vrrp-wan-reconcile now refuses to bless a box whose drop-in is missing: /etc is per-image, and a VyOS upgrade would otherwise silently remove the protection. may-dial is a LEASE, not a flag. ConditionPathExists is evaluated at start only -- it can prevent a dial, never revoke one -- so a reconciler that stops running while its box is demoted would keep the one ISP session for ever. The reconciler renews the lease; a new 5s vrrp-wan-guard revokes it, and only ever revokes. It fired correctly first time: "GUARD: lease stale (81s > 75s)". Also: remove-then-stop on release (the file's absence blocks a NEW start that a concurrent commit would trigger); a flap damper, because two routers that both believe they hold the VIP will both dial and each dial kills the other's session -- against a real ISP that is how an account gets rate-limited; and a guard on `cfg` returning empty under commit-lock contention, which had already produced one spurious "releasing" on a box that needed nothing. GRACE 90 -> 180. accel-ppp's dead-peer budget is lcp-echo-interval(30) x failure(3) = 90s, so the old value sat exactly on the boundary: a hard failover into an AC that does not replace the stale session would fail its own check, shed the VIPs, and leave both routers in FAULT. The sim could not have tested any of this. Both routers now get the identical WAN -- the secondary had none "because two PPPoE clients sharing one credential is a different failure mode than anything production has", which is backwards: that IS production. It also left the pair incomparable, ten NAT rules against none. Safety now comes from resting state, not asymmetry. Three more things the sim was hiding: - the drift check's secondary regex omitted interfaces pppoe/bonding, nat source and protocols failover, so it reported "in sync" for a box with no WAN at all; - the VRRP health-check and transition-script hooks existed on both live VMs and in NEITHER generator -- the mechanism under test was pure undetected drift; - labsim-vyos's only default route was the libvirt-NAT scaffold, so every "the LAN still has internet" verdict on it was answered by eth2 rather than the WAN. --drop-scaffold applied; the earlier DHCP-failover proof is being re-run because of it. vrrp-wan-install ends the other half of that: the sim's previous proof came from scripts hand-`sed`-ed in place, so the tested behaviour was not the committed behaviour. `--check` now makes that a hard failure. First green run: master holds both WANs, backup released, and the AC reports exactly ONE session. sim-net-apply.sh check: all four in sync. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DMVzWZgiKW2wquf5z8S1yH
2026-09-05 18:50:16 +01:00
# No `save`, deliberately. config.boot keeps `vif 53 disable` on BOTH routers, so
# a reboot in any order comes up unable to claim the cloned MAC. PPPoE needs no
# such convention any more: with the gate, config.boot is safe by construction
# and a stray `save` cannot make both boxes dial.