Merge pull request #27330 from winterqt/push-pwouqluolwxt

libpod: fix healthchecks not executing every interval on linux

libpod/healthcheck_linux.go

@@ -40,7 +40,8 @@ func (c *Container) createTimer(interval string, isStartup bool) error {
 		cmd = append(cmd, "--setenv=PATH="+path)
 	}
 
-	cmd = append(cmd, "--unit", hcUnitName, fmt.Sprintf("--on-unit-inactive=%s", interval), "--timer-property=AccuracySec=1s", podman)
+	// StartLimitIntervalSec=0 so we don't hit the restart limit
+	cmd = append(cmd, "--unit", hcUnitName, fmt.Sprintf("--on-unit-inactive=%s", interval), "--timer-property=AccuracySec=1s", "--property=StartLimitIntervalSec=0", podman)
 
 	if logrus.IsLevelEnabled(logrus.DebugLevel) {
 		cmd = append(cmd, "--log-level=debug", "--syslog")

test/system/220-healthcheck.bats

@@ -104,6 +104,11 @@ Log[-1].Output   | \"Uh-oh on stdout!\\\nUh-oh on stderr!\\\n\"
     assert "$cidmatch" =~ " $cid-[0-9a-f]+\.timer  *.*/podman healthcheck run $cid" \
            "Healthcheck systemd unit exists"
 
+    # Check that the right service option is applied so we don't hit the systemd restart limit.
+    # Even though the code sets StartLimitIntervalSec the systemd command prints StartLimitInterval*U*Sec
+    run -0 systemctl show "$cid-*.service"
+    assert "$output" =~ "StartLimitIntervalUSec=0" "The hc service has the right interval set"
+
     current_time=$(date --iso-8601=ns)
     # After three successive failures, container should no longer be healthy
     _check_health $ctrname "Four or more failures" "