feat: Harden fabric-router DaemonSet defaults - #274
Merged
Conversation
Adds livenessProbe (pidof zebra/bgpd), resource requests/limits, a preStop hook for graceful FRR shutdown, and rollingUpdate strategy — all standard, environment-agnostic defaults for an FRR DaemonSet so every consumer (prod, staging, containerlab) gets them without needing a per-environment patch. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Research against Calico/MetalLB/Cilium's production FRR/BGP DaemonSets surfaced several gaps versus real-world practice: - pidof zebra bgpd only proves the processes are resident, not that they're responsive, and passes as long as any one of the named processes is alive. Replace with per-daemon vtysh checks (`vtysh -d <daemon> -c "show version"`) — verified via FRR source that `vtysh -c "show daemons"` silently omits unreachable daemons from its exit code, so it wouldn't have caught a dead bgpd either. - Add a startupProbe so a slow config load (peer groups, route-maps, prefix-lists) doesn't get killed by liveness before it's up. - Add priorityClassName: system-node-critical, matching calico-node and cilium-agent, so the speaker isn't evicted under node pressure. - Add dnsPolicy: ClusterFirstWithHostNet, required alongside hostNetwork: true and missing before now. - Bump cpu request 50m -> 100m to absorb BFD/convergence bursts. terminationGracePeriodSeconds is intentionally left untouched: calico-node and MetalLB's speaker both ship low values (0s and 2s) rather than stretching it to match BGP hold-timers, relying on BFD for fast peer-side failure detection instead of a k8s-side graceful drain. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
ecv
approved these changes
Aug 3, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
The canonical
fabric-routerDaemonSet baseline had no liveness/startup probes, no resource requests/limits, no graceful shutdown hook, and no rolling update strategy — gaps that would otherwise cost every consumer (prod, staging, containerlab) real operational safety once they adopt this baseline.updateStrategy.rollingUpdate.maxUnavailable: 1so rollouts don't take down the underlay fleet-wide at once.preStophook and resource requests/limits sized for FRR's actual footprint at this scale (small peer/route count, not a full-table edge router).vtysh -d <daemon> -c "show version"calls instead ofpidof zebra bgpd— verified against FRR's source that apidof-style orvtysh -c "show daemons"check can't reliably detect a single dead daemon, since it only needs one named process alive (or, forshow daemons, silently omits unreachable daemons without affecting the exit code).startupProbeso a slow config load (peer groups, route-maps, prefix-lists) isn't killed by liveness before it's up.priorityClassName: system-node-criticalanddnsPolicy: ClusterFirstWithHostNet