cmd/relui: address performance issues with module zips

Creating the module zips is surprisingly CPU-intensive, and we end up
doing it many times in parallel, resulting in it taking more than 10
minutes to finish and triggering the watchdog timer.

Disable the watchdog for this task, and give relui more CPU so that it's
less of a problem in the first place. (Collateral damage to other tasks
is still a concern.)

For golang/go#58659.

Change-Id: I65b5c47f9fdc011885c0d5f886f4050551d55f61
Reviewed-on: https://go-review.googlesource.com/c/build/+/489275
Reviewed-by: Dmitri Shuralyov <[email protected]>
Auto-Submit: Heschi Kreinick <[email protected]>
TryBot-Result: Gopher Robot <[email protected]>
Reviewed-by: Dmitri Shuralyov <[email protected]>
Run-TryBot: Heschi Kreinick <[email protected]>

Author: heschi

cmd/relui/deployment-prod.yaml

@@ -59,8 +59,8 @@ spec:
               value: relui
           resources:
             requests:
-              cpu: "2"
-              memory: "4Gi"
+              cpu: "8"
+              memory: "8Gi"
         - name: cloud-sql-proxy
           # It is recommended to use the latest version of the Cloud SQL proxy
           image: gcr.io/cloudsql-docker/gce-proxy:latest

internal/relui/workflows.go

@@ -717,6 +717,7 @@ func (b *BuildReleaseTasks) modFilesFromDistpack(ctx *wf.TaskContext, distpack a
 func (b *BuildReleaseTasks) modFilesFromBinary(ctx *wf.TaskContext, target *releasetargets.Target, version string, t time.Time, tar artifact) (moduleArtifact, error) {
 	result := moduleArtifact{Target: tar.Target}
 	a, err := b.runBuildStep(ctx, nil, nil, tar, "mod.zip", func(_ *task.BuildletStep, r io.Reader, w io.Writer) error {
+		ctx.DisableWatchdog() // The zipping process can be time consuming and is unlikely to hang.
 		var err error
 		result.Mod, result.Info, err = task.TarToModFiles(target, version, t, r, w)
 		return err

internal/workflow/workflow.go

@@ -460,11 +460,20 @@ func (c *TaskContext) DisableRetries() {
 }
 
 func (c *TaskContext) ResetWatchdog() {
+	c.resetWatchdog(WatchdogDelay)
+}
+
+func (c *TaskContext) DisableWatchdog() {
+	// Resetting with a very long delay is easier than canceling the timer.
+	c.resetWatchdog(365 * 24 * time.Hour)
+}
+
+func (c *TaskContext) resetWatchdog(d time.Duration) {
 	// Should only occur in tests.
 	if c.watchdogTimer == nil {
 		return
 	}
-	c.watchdogTimer.Reset(WatchdogDelay)
+	c.watchdogTimer.Reset(d)
 }
 
 // A Listener is used to notify the workflow host of state changes, for display