zoekt-webserver: trigger watchdog after 3 failures
We have observed zoekt-webserver being killed by the watchdog. This typically
can happen on boot were a few other services are also competing for
resources. Used a failure count rather than a longer timeout since that is
usually how I configure the health checks in Kubernetes.
Change-Id: I0698840477bc188768badef25fcd81a89c3ac911
diff --git a/cmd/zoekt-webserver/main.go b/cmd/zoekt-webserver/main.go
index a0ab5d0..46973fa 100644
--- a/cmd/zoekt-webserver/main.go
+++ b/cmd/zoekt-webserver/main.go
@@ -246,9 +246,15 @@
}
tick := time.NewTicker(dt)
+ errCount := 0
for range tick.C {
err := watchdogOnce(context.Background(), client, addr)
if err != nil {
+ errCount++
+ } else {
+ errCount = 0
+ }
+ if errCount == 3 {
log.Panicf("watchdog: %v", err)
}
}
