gsadmin/spinifex
1
0
Fork 0
mirror of https://github.com/mulgadc/spinifex.git synced 2026-10-19 19:01:42 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(vpcd): also subscribe DHCPManager before bootstrap reconcile pass

Browse Source 
Previous fix moved DHCPManager.Subscribe ahead of ReconcileFromKV but
missed Pass 1 (the bootstrap-config Reconcile that runs even earlier).
On a dev box with a source="dhcp" pool and a bootstrap IGW, that earlier
pass still hit "no responders" and burned the 60-retry cold-boot budget
before the Manager came up. Move both reconcile passes after Subscribe.
This commit is contained in:
joshsiv-mulga
2026-04-30 21:58:15 +10:00
parent a6ef8e575b
commit 714898022e
1 changed files with 14 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
spinifex/services/vpcd/vpcd.go
+14 -10
View File
@@ -363,11 +363,6 @@ func launchService(cfg *Config) error {
holder, _ := os.Hostname()
releaseLeader, isLeader := AcquireReconcileLeader(nc, holder)
// Run reconciliation before subscribing (leader only)
if isLeader {
Reconcile(ctx, topo, cfg.Bootstrap)
}
// Macvlan mode only: align macvlan MAC with the OVN gateway router MAC.
// The macvlan only delivers inbound unicast matching its own MAC. With
// centralized NAT, OVN uses the router MAC for all external ARP replies.
@@ -396,11 +391,14 @@ func launchService(cfg *Config) error {
// DHCP manager: services vpc.dhcp.acquire/release requests from the
// daemon-side ExternalIPAM handlers and (mulga-siv-38) topology's gw-LRP
// allocator. MUST subscribe before ReconcileFromKV / retrofit so those
// passes don't fan out vpc.dhcp.acquire requests to "no responders" and
// burn the 60-attempt cold-boot retry budget per existing IGW. Started
// unconditionally — pools with source="static" never issue acquire
// requests, so the Manager sits idle until something actually wants DHCP.
// allocator. MUST subscribe before any reconcile pass — bootstrap
// reconcile, ReconcileFromKV, and retrofit all call into reconcileIGW /
// expectedGatewayPortNetwork which fan out vpc.dhcp.acquire on
// source="dhcp" pools. Without a live subscription those requests hit
// "no responders" and burn the 60-attempt cold-boot retry budget per
// existing IGW. Started unconditionally — pools with source="static"
// never issue acquire requests, so the Manager sits idle until
// something actually wants DHCP.
js, err := nc.JetStream()
if err != nil {
slog.Error("Failed to get JetStream context for DHCP manager", "err", err)
@@ -428,6 +426,12 @@ func launchService(cfg *Config) error {
}
}()
// Pass 1: Bootstrap-config reconcile. Creates the bootstrap VPC / IGW
// topology if missing. Leader-only.
if isLeader {
Reconcile(ctx, topo, cfg.Bootstrap)
}
// Pass 2: Reconcile from NATS KV (handles reboots, OVN DB loss, missed events).
// Runs after subscribing so new events are not missed during reconciliation.
// Leader-gated (mulga-siv-29) to avoid concurrent Create races against OVN NB.