cmd/worker: catch when the server is about to get killed
And log how long it was alive. This is achieved by catching a SIGTERM
signal. Cloud Run shuts down a container 10s after this signal is sent,
so this allows us to print the state of the server. The hope is that
we can find out why our instance are being killed often.
https://cloud.google.com/run/docs/container-contract#lifecycle
Change-Id: I6df4a96843cf4e82e2c485464afbf440263abcb7
Reviewed-on: https://go-review.googlesource.com/c/pkgsite-metrics/+/539775
LUCI-TryBot-Result: Go LUCI <golang-scoped@luci-project-accounts.iam.gserviceaccount.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Run-TryBot: Zvonimir Pavlinovic <zpavlinovic@google.com>
Reviewed-by: Jonathan Amsterdam <jba@google.com>
diff --git a/cmd/worker/main.go b/cmd/worker/main.go
index 6c9b212..bca9f99 100644
--- a/cmd/worker/main.go
+++ b/cmd/worker/main.go
@@ -11,6 +11,8 @@
"fmt"
"net/http"
"os"
+ "os/signal"
+ "syscall"
"time"
"golang.org/x/exp/slog"
@@ -68,10 +70,24 @@
cfg.Insecure = *insecure
cfg.Dump(os.Stdout)
log.Infof(ctx, "config: project=%s, dataset=%s", cfg.ProjectID, cfg.BigQueryDataset)
+
if _, err := worker.NewServer(ctx, cfg); err != nil {
return err
}
+ go monitor(ctx)
+
addr := ":" + *port
log.Infof(ctx, "Listening on addr http://localhost%s", addr)
return fmt.Errorf("listening: %v", http.ListenAndServe(addr, nil))
}
+
+// monitor measures details of server execution from
+// the moment is starts listening to the moment it
+// gets a SIGTERM signal.
+func monitor(ctx context.Context) {
+ start := time.Now()
+ signals := make(chan os.Signal, 1)
+ signal.Notify(signals, syscall.SIGTERM)
+ <-signals
+ log.Infof(ctx, "server stopped listening after: %v\n", time.Since(start))
+}
