test: Log command output at fail level if action failed

Log expected command failure at info level so that it is easier to
reason about the following flow logs, if any.

Always consider any stderr output as a failure and log it at fail
level.

Log stdout at fail level if the action failed, and at the debug level
otherwise.

This helps diagnoze test failures also when `--debug` is not on.

Signed-off-by: Jarno Rajahalme <[email protected]>

connectivity/check/action.go

@@ -167,23 +167,32 @@ func (a *Action) ExecInPod(ctx context.Context, cmd []string) {
 	cmdName := cmd[0]
 	cmdStr := strings.Join(cmd, " ")
 
-	if stderr.Len() > 0 {
-		a.test.Debugf("%s stderr: %s", cmdName, stderr.String())
-	} else if stdout.Len() > 0 {
-		a.test.Debugf("%s stdout: %s", cmdName, stdout.String())
-	}
-
 	if err != nil || stderr.Len() > 0 {
 		if a.shouldSucceed() {
 			a.Failf("command %q failed: %s", cmdStr, err)
 		} else {
-			a.test.Debugf("command %q failed as expected: %s", cmdStr, err)
+			a.test.Infof("command %q failed as expected: %s", cmdStr, err)
 		}
 	} else {
 		if !a.shouldSucceed() {
 			a.Failf("command %q succeeded while it should have failed: %s", cmdStr, stdout.String())
 		}
 	}
+	if stderr.Len() > 0 {
+		// Any stderr output is considered a test failure.
+		a.Failf("%s stderr: %s", cmdName, stderr.String())
+	}
+	if stdout.Len() > 0 {
+		// Debug level logs are only shown when running with `--debug`,
+		// which is proper for succeeding actions. Log at the fail
+		// level when the action has failed to always log the failing
+		// output.
+		if a.failed {
+			a.Failf("%s stdout: %s", cmdName, stdout.String())
+		} else {
+			a.Debugf("%s stdout: %s", cmdName, stdout.String())
+		}
+	}
 }
 
 // shouldSucceed returns true if no drops are expected in either direction.