Merge pull request #26515 from dashea/dshea-3.2.3-healthcheckbackports

[v3.2.3-rhel] Backports for health check files

contrib/cirrus/setup_environment.sh

@@ -19,7 +19,7 @@ die_unknown() {
     die "Unknown/unsupported \$$var_name '$var_value'"
 }
 
-_EOL=20250531
+_EOL=20270922
 if [[ $(date +%Y%m%d) -ge $_EOL ]]; then
     die "As of $_EOL this branch is probably
 no longer supported in RHEL 8.4.z, please

libpod/container_internal.go

@@ -1765,7 +1765,7 @@ func (c *Container) cleanupStorage() error {
 	return cleanupErr
 }
 
-// Unmount the a container and free its resources
+// Unmount the container and free its resources
 func (c *Container) cleanup(ctx context.Context) error {
 	var lastError error
 
@@ -1773,8 +1773,8 @@ func (c *Container) cleanup(ctx context.Context) error {
 
 	// Remove healthcheck unit/timer file if it execs
 	if c.config.HealthCheckConfig != nil {
-		if err := c.removeTimer(); err != nil {
-			logrus.Errorf("Error removing timer for container %s healthcheck: %v", c.ID(), err)
+		if err := c.removeTransientFiles(ctx); err != nil {
+			logrus.Errorf("Removing timer for container %s healthcheck: %v", c.ID(), err)
 		}
 	}
 

libpod/healthcheck_linux.go

@@ -1,6 +1,7 @@
 package libpod
 
 import (
+	"context"
 	"fmt"
 	"os"
 	"os/exec"
@@ -59,9 +60,9 @@ func (c *Container) startTimer() error {
 	return err
 }
 
-// removeTimer removes the systemd timer and unit files
+// removeTransientFiles removes the systemd timer and unit files
 // for the container
-func (c *Container) removeTimer() error {
+func (c *Container) removeTransientFiles(ctx context.Context) error {
 	if c.disableHealthCheckSystemd() {
 		return nil
 	}
@@ -71,12 +72,29 @@ func (c *Container) removeTimer() error {
 	}
 	defer conn.Close()
 	timerFile := fmt.Sprintf("%s.timer", c.ID())
-	_, err = conn.StopUnit(timerFile, "fail", nil)
+	serviceFile := fmt.Sprintf("%s.service", c.ID())
 
-	// We want to ignore errors where the timer unit has already been removed. The error
-	// return is generic so we have to check against the string in the error
-	if err != nil && strings.HasSuffix(err.Error(), ".timer not loaded.") {
-		return nil
+	// If the service has failed (the healthcheck has failed), then
+	// the .service file is not removed on stopping the unit file. If
+	// we check the properties of the service, it will automatically
+	// reset the state.  But checking the state takes msecs vs usecs to
+	// blindly call reset.
+	if err := conn.ResetFailedUnitContext(ctx, serviceFile); err != nil {
+		logrus.Debugf("failed to reset unit file: %q", err)
+	}
+
+	// We want to ignore errors where the timer unit and/or service unit has already
+	// been removed. The error return is generic so we have to check against the
+	// string in the error
+	if _, err = conn.StopUnitContext(ctx, serviceFile, "fail", nil); err != nil {
+		if !strings.HasSuffix(err.Error(), ".service not loaded.") {
+			return errors.Wrapf(err, "unable to remove service file")
+		}
+	}
+	if _, err = conn.StopUnitContext(ctx, timerFile, "fail", nil); err != nil {
+		if strings.HasSuffix(err.Error(), ".timer not loaded.") {
+			return nil
+		}
 	}
 	return err
 }

test/e2e/healthcheck_run_test.go

@@ -8,6 +8,7 @@ import (
 	. "github.com/containers/podman/v3/test/utils"
 	. "github.com/onsi/ginkgo"
 	. "github.com/onsi/gomega"
+	. "github.com/onsi/gomega/gexec"
 )
 
 var _ = Describe("Podman healthcheck run", func() {
@@ -203,4 +204,20 @@ var _ = Describe("Podman healthcheck run", func() {
 		inspect = podmanTest.InspectContainer("hc")
 		Expect(inspect[0].State.Healthcheck.Status).To(Equal("healthy"))
 	})
+
+	It("stopping and then starting a container with healthcheck cmd", func() {
+		session := podmanTest.Podman([]string{"run", "-dt", "--name", "hc", "--health-cmd", "[\"ls\", \"/foo\"]", ALPINE, "top"})
+		session.WaitWithDefaultTimeout()
+		Expect(session).Should(Exit(0))
+
+		stop := podmanTest.Podman([]string{"stop", "-t0", "hc"})
+		stop.WaitWithDefaultTimeout()
+		Expect(stop).Should(Exit(0))
+
+		startAgain := podmanTest.Podman([]string{"start", "hc"})
+		startAgain.WaitWithDefaultTimeout()
+		Expect(startAgain).Should(Exit(0))
+		Expect(startAgain.OutputToString()).To(Equal("hc"))
+		Expect(startAgain.ErrorToString()).To(Equal(""))
+	})
 })