Merge pull request #12359 from hashicorp/dnephin/fix-debug-size

debug: limit the size of the trace

.changelog/12359.txt

@@ -0,0 +1,3 @@
+```release-note:improvement
+debug: reduce the capture time for trace to only a single interval instead of the full duration to make trace.out easier to open without running into OOM errors.
+```

command/debug/debug.go

@@ -400,20 +400,16 @@ func makeIntervalDir(base string, now time.Time) (string, error) {
 
 func (c *cmd) captureLongRunning(ctx context.Context) error {
 	g := new(errgroup.Group)
-	// Capture a profile/trace with a minimum of 1s
-	s := c.duration.Seconds()
-	if s < 1 {
-		s = 1
-	}
+
 	if c.captureTarget(targetProfiles) {
 		g.Go(func() error {
 			// use ctx without a timeout to allow the profile to finish sending
-			return c.captureProfile(ctx, s)
+			return c.captureProfile(ctx, c.duration.Seconds())
 		})
 
 		g.Go(func() error {
 			// use ctx without a timeout to allow the trace to finish sending
-			return c.captureTrace(ctx, s)
+			return c.captureTrace(ctx, int(c.interval.Seconds()))
 		})
 	}
 	if c.captureTarget(targetLogs) {
@@ -443,8 +439,8 @@ func (c *cmd) captureGoRoutines(outputDir string) error {
 	return ioutil.WriteFile(filepath.Join(outputDir, "goroutine.prof"), gr, 0644)
 }
 
-func (c *cmd) captureTrace(ctx context.Context, s float64) error {
-	prof, err := c.client.Debug().PProf(ctx, "trace", int(s))
+func (c *cmd) captureTrace(ctx context.Context, duration int) error {
+	prof, err := c.client.Debug().PProf(ctx, "trace", duration)
 	if err != nil {
 		return fmt.Errorf("failed to collect cpu profile: %w", err)
 	}

website/content/commands/debug.mdx

@@ -78,7 +78,7 @@ information when `debug` is running. By default, it captures all information.
 | `members` | A list of all the WAN and LAN members in the cluster.                                                                                                                                                                                                                                                                                                                                                                     |
 | `metrics` | Metrics from the in-memory metrics endpoint in the target, captured at the interval.                                                                                                                                                                                                                                                                                                                                      |
 | `logs`    | `DEBUG` level logs for the target agent, captured for the duration.                                                                                                                                                                                                                                                                                                                                                       |
-| `pprof`   | Golang heap, CPU, goroutine, and trace profiling. CPU and traces are captured for `duration` in a single file while heap and goroutine are separate snapshots for each `interval`. This information is not retrieved unless [`enable_debug`](/docs/agent/options#enable_debug) is set to `true` on the target agent or ACLs are enable and an ACL token with `operator:read` is provided. |
+| `pprof`   | Golang heap, CPU, goroutine, and trace profiling. CPU profile is captured for `duration` in a single file, trace is captured for a single `interval`, while heap and goroutine are separate snapshots for each `interval`. This information is not retrieved unless [`enable_debug`](/docs/agent/options#enable_debug) is set to `true` on the target agent or ACLs are enable and an ACL token with `operator:read` is provided. |
 
 ## Examples