Added retry for snapshot and s3 upload, verify etcd running on host (#2952)

Added retry for snapshot and s3 upload, verify etcd running on host

Added option to quiet noisy container logs

cluster/addons.go

@@ -557,7 +557,6 @@ func (c *Cluster) StoreAddonConfigMap(ctx context.Context, addonYaml string, add
 	timeout := make(chan bool, 1)
 	go func() {
 		for {
-
 			updated, err = k8s.UpdateConfigMap(kubeClient, []byte(addonYaml), addonName)
 			if err != nil {
 				time.Sleep(time.Second * 5)

cluster/etcd.go

@@ -3,6 +3,7 @@ package cluster
 import (
 	"context"
 	"fmt"
+	"strings"
 
 	"github.com/sirupsen/logrus"
 
@@ -17,16 +18,52 @@ import (
 
 func (c *Cluster) SnapshotEtcd(ctx context.Context, snapshotName string) error {
 	backupImage := c.getBackupImage()
-	for _, host := range c.EtcdHosts {
 	containerTimeout := DefaultEtcdBackupConfigTimeout
 	if c.Services.Etcd.BackupConfig != nil && c.Services.Etcd.BackupConfig.Timeout > 0 {
 		containerTimeout = c.Services.Etcd.BackupConfig.Timeout
 	}
+
+	// store first error message
+	var snapshotErr error
+	snapshotFailures := 0
+	s3UploadFailures := 0
+
+	for _, host := range c.EtcdHosts {
 		newCtx := context.WithValue(ctx, docker.WaitTimeoutContextKey, containerTimeout)
 		if err := services.RunEtcdSnapshotSave(newCtx, host, c.PrivateRegistriesMap, backupImage, snapshotName, true, c.Services.Etcd, c.Version); err != nil {
-			return err
+			if strings.Contains(err.Error(), "failed to upload etcd snapshot file to s3 on host") {
+				s3UploadFailures++
+			} else {
+				if snapshotErr == nil {
+					snapshotErr = err
+				}
+				snapshotFailures++
 			}
 		}
+	}
+
+	if snapshotFailures == len(c.EtcdHosts) {
+		log.Warnf(ctx, "[etcd] Failed to take snapshot on all etcd hosts: %s", snapshotErr)
+		return fmt.Errorf("[etcd] Failed to take snapshot on all etcd hosts: %s", snapshotErr)
+	} else if snapshotFailures > 0 {
+		log.Warnf(ctx, "[etcd] Failed to take snapshot on %s etcd hosts", snapshotFailures)
+	} else {
+		log.Infof(ctx, "[etcd] Finished saving snapshot [%s] on all etcd hosts", snapshotName)
+	}
+
+	if c.Services.Etcd.BackupConfig.S3BackupConfig == nil {
+		return nil
+	}
+
+	if s3UploadFailures >= len(c.EtcdHosts)-snapshotFailures {
+		log.Warnf(ctx, "[etcd] Failed to upload etcd snapshot file to s3 on all etcd hosts")
+		return fmt.Errorf("[etcd] Failed to upload etcd snapshot file to s3 on all etcd hosts")
+	} else if s3UploadFailures > 0 {
+		log.Warnf(ctx, "[etcd] Failed to upload etcd snapshot file to s3 on %s etcd hosts", s3UploadFailures)
+	} else {
+		log.Infof(ctx, "[etcd] Finished uploading etcd snapshot file to s3 on all etcd hosts")
+	}
+
 	return nil
 }
 
@@ -112,13 +149,28 @@ func (c *Cluster) PrepareBackup(ctx context.Context, snapshotPath string) error
 	var backupServer *hosts.Host
 	backupImage := c.getBackupImage()
 	var errors []error
-	if c.Services.Etcd.BackupConfig == nil || // legacy rke local backup
+	// s3 backup case
-		(c.Services.Etcd.BackupConfig != nil && c.Services.Etcd.BackupConfig.S3BackupConfig == nil) { // rancher local backup
+	if c.Services.Etcd.BackupConfig != nil &&
+		c.Services.Etcd.BackupConfig.S3BackupConfig != nil {
+		log.Infof(ctx, "[etcd] etcd s3 backup configuration found, will use s3 as source")
+		downloadFailed := false
+		for _, host := range c.EtcdHosts {
+			if err := services.DownloadEtcdSnapshotFromS3(ctx, host, c.PrivateRegistriesMap, backupImage, snapshotPath, c.Services.Etcd, c.Version); err != nil {
+				log.Warnf(ctx, "failed to download snapshot [%s] from s3 on host [%s]: %v", snapshotPath, host.Address, err)
+				downloadFailed = true
+				break
+			}
+		}
+		backupReady = !downloadFailed
+	}
+	// legacy rke local backup or rancher local backup
+	if !backupReady {
 		if c.Services.Etcd.BackupConfig == nil {
 			log.Infof(ctx, "[etcd] No etcd snapshot configuration found, will use local as source")
-		}
+		} else if c.Services.Etcd.BackupConfig.S3BackupConfig == nil {
-		if c.Services.Etcd.BackupConfig != nil && c.Services.Etcd.BackupConfig.S3BackupConfig == nil {
 			log.Infof(ctx, "[etcd] etcd snapshot configuration found and no s3 backup configuration found, will use local as source")
+		} else {
+			log.Warnf(ctx, "[etcd] etcd snapshot configuration found and s3 backup configuration failed, falling back to use local as source")
 		}
 		// stop etcd on all etcd nodes, we need this because we start the backup server on the same port
 		for _, host := range c.EtcdHosts {
@@ -159,17 +211,6 @@ func (c *Cluster) PrepareBackup(ctx context.Context, snapshotPath string) error
 		backupReady = true
 	}
 
-	// s3 backup case
-	if c.Services.Etcd.BackupConfig != nil &&
-		c.Services.Etcd.BackupConfig.S3BackupConfig != nil {
-		log.Infof(ctx, "[etcd] etcd s3 backup configuration found, will use s3 as source")
-		for _, host := range c.EtcdHosts {
-			if err := services.DownloadEtcdSnapshotFromS3(ctx, host, c.PrivateRegistriesMap, backupImage, snapshotPath, c.Services.Etcd, c.Version); err != nil {
-				return err
-			}
-		}
-		backupReady = true
-	}
 	if !backupReady {
 		return fmt.Errorf("failed to prepare backup for restore")
 	}

docker/docker.go

@@ -124,7 +124,7 @@ func DoRunOnetimeContainer(ctx context.Context, dClient *client.Client, imageCfg
 		}
 		log.Infof(ctx, "Successfully started [%s] container on host [%s]", containerName, hostname)
 		log.Infof(ctx, "Waiting for [%s] container to exit on host [%s]", containerName, hostname)
-		exitCode, err := WaitForContainer(ctx, dClient, hostname, containerName)
+		exitCode, err := WaitForContainer(ctx, dClient, hostname, containerName, true)
 		if err != nil {
 			return fmt.Errorf("Container [%s] did not complete in time on host [%s]", containerName, hostname)
 		}
@@ -164,11 +164,11 @@ func DoRollingUpdateContainer(ctx context.Context, dClient *client.Client, image
 		return fmt.Errorf("[%s] Failed rolling update of container: docker client is nil for container [%s] on host [%s]", plane, containerName, hostname)
 	}
 	logrus.Debugf("[%s] Checking for deployed [%s]", plane, containerName)
-	isRunning, err := IsContainerRunning(ctx, dClient, hostname, containerName, false)
+	exists, err := DoesContainerExist(ctx, dClient, hostname, containerName, false)
 	if err != nil {
 		return err
 	}
-	if !isRunning {
+	if !exists {
 		logrus.Debugf("[%s] Container %s is not running on host [%s]", plane, containerName, hostname)
 		return nil
 	}
@@ -217,30 +217,52 @@ func DoRemoveContainer(ctx context.Context, dClient *client.Client, containerNam
 	return nil
 }
 
-func IsContainerRunning(ctx context.Context, dClient *client.Client, hostname string, containerName string, all bool) (bool, error) {
+func FindContainer(ctx context.Context, dClient *client.Client, hostname string, containerName string, all bool) (*types.Container, error) {
 	if dClient == nil {
-		return false, fmt.Errorf("Failed to check if container is running: docker client is nil for container [%s] on host [%s]", containerName, hostname)
+		return nil, fmt.Errorf("Failed to find container: docker client is nil for container [%s] on host [%s]", containerName, hostname)
 	}
 	var containers []types.Container
 	var err error
 	for i := 1; i <= RetryCount; i++ {
-		logrus.Infof("Checking if container [%s] is running on host [%s], try #%d", containerName, hostname, i)
+		logrus.Infof("Finding container [%s] on host [%s], try #%d", containerName, hostname, i)
 		containers, err = dClient.ContainerList(ctx, types.ContainerListOptions{All: all})
 		if err != nil {
-			logrus.Warnf("Error checking if container [%s] is running on host [%s]: %v", containerName, hostname, err)
+			logrus.Warnf("Error finding container [%s] exists on host [%s]: %v", containerName, hostname, err)
 			continue
 		}
 		break
 	}
 	if err != nil {
-		return false, fmt.Errorf("Error checking if container [%s] is running on host [%s]: %v", containerName, hostname, err)
+		return nil, fmt.Errorf("Error checking if container [%s] exists on host [%s]: %v", containerName, hostname, err)
 	}
 	for _, container := range containers {
 		if len(container.Names) != 0 && container.Names[0] == "/"+containerName {
-			return true, nil
+			return &container, nil
 		}
 	}
-	return false, nil
+	return nil, nil
+}
+
+func DoesContainerExist(ctx context.Context, dClient *client.Client, hostname string, containerName string, all bool) (bool, error) {
+	if dClient == nil {
+		return false, fmt.Errorf("Failed to check if container exists: docker client is nil for container [%s] on host [%s]", containerName, hostname)
+	}
+	container, err := FindContainer(ctx, dClient, hostname, containerName, all)
+	if err != nil {
+		return false, fmt.Errorf("Error checking if container [%s] is running on host [%s]: %v", containerName, hostname, err)
+	}
+	return container != nil, nil
+}
+
+func IsContainerRunning(ctx context.Context, dClient *client.Client, hostname string, containerName string, all bool) (bool, error) {
+	if dClient == nil {
+		return false, fmt.Errorf("Failed to check if container is running: docker client is nil for container [%s] on host [%s]", containerName, hostname)
+	}
+	container, err := FindContainer(ctx, dClient, hostname, containerName, all)
+	if err != nil {
+		return false, fmt.Errorf("Error checking if container [%s] is running on host [%s]: %v", containerName, hostname, err)
+	}
+	return container != nil && container.State == "running", nil
 }
 
 func localImageExists(ctx context.Context, dClient *client.Client, hostname string, containerImage string) error {
@@ -476,7 +498,7 @@ func StopRenameContainer(ctx context.Context, dClient *client.Client, hostname s
 		return fmt.Errorf("Failed to stop and rename container: docker client is nil for container [%s] on host [%s]", oldContainerName, hostname)
 	}
 	// make sure we don't have an old old-container from a previous broken update
-	exists, err := IsContainerRunning(ctx, dClient, hostname, newContainerName, true)
+	exists, err := DoesContainerExist(ctx, dClient, hostname, newContainerName, true)
 	if err != nil {
 		return err
 	}
@@ -488,14 +510,14 @@ func StopRenameContainer(ctx context.Context, dClient *client.Client, hostname s
 	if err := StopContainer(ctx, dClient, hostname, oldContainerName); err != nil {
 		return err
 	}
-	if _, err := WaitForContainer(ctx, dClient, hostname, oldContainerName); err != nil {
+	if _, err := WaitForContainer(ctx, dClient, hostname, oldContainerName, true); err != nil {
 		return err
 	}
 	return RenameContainer(ctx, dClient, hostname, oldContainerName, newContainerName)
 
 }
 
-func WaitForContainer(ctx context.Context, dClient *client.Client, hostname string, containerName string) (int64, error) {
+func WaitForContainer(ctx context.Context, dClient *client.Client, hostname string, containerName string, noisy bool) (int64, error) {
 	if dClient == nil {
 		return 1, fmt.Errorf("Failed waiting for container: docker client is nil for container [%s] on host [%s]", containerName, hostname)
 	}
@@ -504,8 +526,9 @@ func WaitForContainer(ctx context.Context, dClient *client.Client, hostname stri
 	if v, ok := ctx.Value(WaitTimeoutContextKey).(int); ok && v > 0 {
 		containerTimeout = v
 	}
-	for retries := 0; retries < containerTimeout; retries++ {
 	log.Infof(ctx, "Waiting for [%s] container to exit on host [%s]", containerName, hostname)
+	var lastStdout, lastStderr string
+	for retries := 0; retries < containerTimeout; retries++ {
 		container, err := InspectContainer(ctx, dClient, hostname, containerName)
 		if err != nil {
 			return 1, fmt.Errorf("Could not inspect container [%s] on host [%s]: %s", containerName, hostname, err)
@@ -515,8 +538,12 @@ func WaitForContainer(ctx context.Context, dClient *client.Client, hostname stri
 			if err != nil {
 				logrus.Warnf("Failed to get container logs from container [%s] on host [%s]: %v", containerName, hostname, err)
 			}
-
+			if noisy || lastStdout != stdout || lastStderr != stderr {
 				log.Infof(ctx, "Container [%s] is still running on host [%s]: stderr: [%s], stdout: [%s]", containerName, hostname, stderr, stdout)
+				lastStdout = stdout
+				lastStderr = stderr
+			}
+
 			time.Sleep(1 * time.Second)
 			continue
 		}
@@ -789,11 +816,11 @@ func DoRestartContainer(ctx context.Context, dClient *client.Client, containerNa
 	return nil
 }
 
-func GetContainerOutput(ctx context.Context, dClient *client.Client, containerName, hostname string) (int64, string, string, error) {
+func GetContainerOutput(ctx context.Context, dClient *client.Client, containerName, hostname string, noisy bool) (int64, string, string, error) {
 	if dClient == nil {
 		return 1, "", "", fmt.Errorf("Failed to get container output: docker client is nil for container [%s] on host [%s]", containerName, hostname)
 	}
-	status, err := WaitForContainer(ctx, dClient, hostname, containerName)
+	status, err := WaitForContainer(ctx, dClient, hostname, containerName, noisy)
 	if err != nil {
 		return 1, "", "", err
 	}

hosts/hosts.go

@@ -135,7 +135,7 @@ func (h *Host) CleanUp(ctx context.Context, toCleanPaths []string, cleanerImage
 		return err
 	}
 
-	if _, err := docker.WaitForContainer(ctx, h.DClient, h.Address, CleanerContainerName); err != nil {
+	if _, err := docker.WaitForContainer(ctx, h.DClient, h.Address, CleanerContainerName, true); err != nil {
 		return err
 	}
 

pki/deploy.go

@@ -116,7 +116,7 @@ func DeployStateOnPlaneHost(ctx context.Context, host *hosts.Host, stateDownload
 		logrus.Warnf("[state] Error during copying state file [%s] to node [%s]: %v", stateFilePath, host.Address, err)
 	}
 
-	if _, err := docker.WaitForContainer(ctx, host.DClient, host.Address, StateDeployerContainerName); err != nil {
+	if _, err := docker.WaitForContainer(ctx, host.DClient, host.Address, StateDeployerContainerName, true); err != nil {
 		return err
 	}
 
@@ -125,12 +125,11 @@ func DeployStateOnPlaneHost(ctx context.Context, host *hosts.Host, stateDownload
 
 func doRunDeployer(ctx context.Context, host *hosts.Host, containerEnv []string, certDownloaderImage string, prsMap map[string]v3.PrivateRegistry, k8sVersion string) error {
 	// remove existing container. Only way it's still here is if previous deployment failed
-	isRunning := false
+	exists, err := docker.DoesContainerExist(ctx, host.DClient, host.Address, CrtDownloaderContainer, true)
-	isRunning, err := docker.IsContainerRunning(ctx, host.DClient, host.Address, CrtDownloaderContainer, true)
 	if err != nil {
 		return err
 	}
-	if isRunning {
+	if exists {
 		if err := docker.RemoveContainer(ctx, host.DClient, host.Address, CrtDownloaderContainer); err != nil {
 			return err
 		}
@@ -188,7 +187,7 @@ func doRunDeployer(ctx context.Context, host *hosts.Host, containerEnv []string,
 	}
 	logrus.Debugf("[certificates] Successfully started Certificate deployer container: %s", CrtDownloaderContainer)
 	for {
-		isDeployerRunning, err := docker.IsContainerRunning(ctx, host.DClient, host.Address, CrtDownloaderContainer, false)
+		isDeployerRunning, err := docker.DoesContainerExist(ctx, host.DClient, host.Address, CrtDownloaderContainer, false)
 		if err != nil {
 			return err
 		}
@@ -331,11 +330,11 @@ func FetchFileFromHost(ctx context.Context, filePath, image string, host *hosts.
 		Binds:      Binds,
 		Privileged: true,
 	}
-	isRunning, err := docker.IsContainerRunning(ctx, host.DClient, host.Address, containerName, true)
+	exists, err := docker.DoesContainerExist(ctx, host.DClient, host.Address, containerName, true)
 	if err != nil {
 		return "", err
 	}
-	if !isRunning {
+	if !exists {
 		if err := docker.DoRunContainer(ctx, host.DClient, imageCfg, hostCfg, containerName, host.Address, state, prsMap); err != nil {
 			return "", err
 		}

pki/pki.go

@@ -131,7 +131,7 @@ func SaveBackupBundleOnHost(ctx context.Context, host *hosts.Host, alpineSystemI
 	if err := docker.DoRunContainer(ctx, host.DClient, imageCfg, hostCfg, BundleCertContainer, host.Address, "certificates", prsMap); err != nil {
 		return err
 	}
-	status, err := docker.WaitForContainer(ctx, host.DClient, host.Address, BundleCertContainer)
+	status, err := docker.WaitForContainer(ctx, host.DClient, host.Address, BundleCertContainer, true)
 	if err != nil {
 		return err
 	}

services/etcd.go

@@ -439,7 +439,6 @@ func RunEtcdSnapshotSave(ctx context.Context, etcdHost *hosts.Host, prsMap map[s
 		if hosts.IsDockerSELinuxEnabled(etcdHost) {
 			hostCfg.SecurityOpt = append(hostCfg.SecurityOpt, SELinuxLabel)
 		}
-
 	}
 	hostCfg.Binds = binds
 
@@ -449,18 +448,30 @@ func RunEtcdSnapshotSave(ctx context.Context, etcdHost *hosts.Host, prsMap map[s
 		if err := docker.DoRemoveContainer(ctx, etcdHost.DClient, EtcdSnapshotOnceContainerName, etcdHost.Address); err != nil {
 			return err
 		}
+
+		// If the etcd container is not running the snapshot will never succeed
+		log.Debugf(ctx, "[etcd] Checking if etcd is running on host [%s]", etcdHost.Address)
+		if running, err := docker.IsContainerRunning(ctx, etcdHost.DClient, etcdHost.Address, "etcd", true); err != nil {
+			return err
+		} else if !running {
+			return fmt.Errorf("etcd is not running on host [%s]", etcdHost.Address)
+		}
+
 		if err := docker.DoRunContainer(ctx, etcdHost.DClient, imageCfg, hostCfg, EtcdSnapshotOnceContainerName, etcdHost.Address, ETCDRole, prsMap); err != nil {
 			return err
 		}
-		status, _, stderr, err := docker.GetContainerOutput(ctx, etcdHost.DClient, EtcdSnapshotOnceContainerName, etcdHost.Address)
+		status, _, stderr, err := docker.GetContainerOutput(ctx, etcdHost.DClient, EtcdSnapshotOnceContainerName, etcdHost.Address, false)
 		if status != 0 || err != nil {
 			if removeErr := docker.RemoveContainer(ctx, etcdHost.DClient, etcdHost.Address, EtcdSnapshotOnceContainerName); removeErr != nil {
-				log.Warnf(ctx, "[etcd] Failed to remove container [%s] on host [%s]: %v", removeErr, etcdHost.Address)
+				log.Warnf(ctx, "[etcd] Failed to remove container [%s] on host [%s]: %v", EtcdSnapshotOnceContainerName, removeErr, etcdHost.Address)
 			}
 			if err != nil {
 				return err
 			}
-			return fmt.Errorf("[etcd] Failed to take one-time snapshot on host [%s], exit code [%d]: %v", etcdHost.Address, status, stderr)
+			if strings.Contains(stderr, "failed to upload etcd snapshot file") {
+				return fmt.Errorf("failed to upload etcd snapshot file to s3 on host [%s], exit code [%d]: %v", etcdHost.Address, status, stderr)
+			}
+			return fmt.Errorf("failed to take one-time snapshot on host [%s], exit code [%d]: %v", etcdHost.Address, status, stderr)
 		}
 
 		return docker.RemoveContainer(ctx, etcdHost.DClient, etcdHost.Address, EtcdSnapshotOnceContainerName)
@@ -618,7 +629,7 @@ func DownloadEtcdSnapshotFromS3(ctx context.Context, etcdHost *hosts.Host, prsMa
 		return err
 	}
 
-	status, _, stderr, err := docker.GetContainerOutput(ctx, etcdHost.DClient, EtcdDownloadBackupContainerName, etcdHost.Address)
+	status, _, stderr, err := docker.GetContainerOutput(ctx, etcdHost.DClient, EtcdDownloadBackupContainerName, etcdHost.Address, true)
 	if status != 0 || err != nil {
 		if removeErr := docker.RemoveContainer(ctx, etcdHost.DClient, etcdHost.Address, EtcdDownloadBackupContainerName); removeErr != nil {
 			log.Warnf(ctx, "Failed to remove container [%s]: %v", removeErr)
@@ -689,7 +700,7 @@ func RestoreEtcdSnapshot(ctx context.Context, etcdHost *hosts.Host, prsMap map[s
 	if err := docker.DoRunContainer(ctx, etcdHost.DClient, imageCfg, hostCfg, EtcdRestoreContainerName, etcdHost.Address, ETCDRole, prsMap); err != nil {
 		return err
 	}
-	status, err := docker.WaitForContainer(ctx, etcdHost.DClient, etcdHost.Address, EtcdRestoreContainerName)
+	status, err := docker.WaitForContainer(ctx, etcdHost.DClient, etcdHost.Address, EtcdRestoreContainerName, false)
 	if err != nil {
 		return err
 	}
@@ -779,7 +790,7 @@ func RunEtcdSnapshotRemove(ctx context.Context, etcdHost *hosts.Host, prsMap map
 	if err := docker.DoRunContainer(ctx, etcdHost.DClient, imageCfg, hostCfg, EtcdSnapshotRemoveContainerName, etcdHost.Address, ETCDRole, prsMap); err != nil {
 		return err
 	}
-	status, _, stderr, err := docker.GetContainerOutput(ctx, etcdHost.DClient, EtcdSnapshotRemoveContainerName, etcdHost.Address)
+	status, _, stderr, err := docker.GetContainerOutput(ctx, etcdHost.DClient, EtcdSnapshotRemoveContainerName, etcdHost.Address, true)
 	if status != 0 || err != nil {
 		if removeErr := docker.RemoveContainer(ctx, etcdHost.DClient, etcdHost.Address, EtcdSnapshotRemoveContainerName); removeErr != nil {
 			log.Warnf(ctx, "Failed to remove container [%s]: %v", removeErr)
@@ -831,7 +842,7 @@ func GetEtcdSnapshotChecksum(ctx context.Context, etcdHost *hosts.Host, prsMap m
 	if err := docker.DoRunContainer(ctx, etcdHost.DClient, imageCfg, hostCfg, EtcdChecksumContainerName, etcdHost.Address, ETCDRole, prsMap); err != nil {
 		return checksum, err
 	}
-	if _, err := docker.WaitForContainer(ctx, etcdHost.DClient, etcdHost.Address, EtcdChecksumContainerName); err != nil {
+	if _, err := docker.WaitForContainer(ctx, etcdHost.DClient, etcdHost.Address, EtcdChecksumContainerName, true); err != nil {
 		return checksum, err
 	}
 	stderr, checksum, err = docker.GetContainerLogsStdoutStderr(ctx, etcdHost.DClient, EtcdChecksumContainerName, "1", false)
@@ -999,7 +1010,7 @@ func DownloadEtcdSnapshotFromBackupServer(ctx context.Context, etcdHost *hosts.H
 		return err
 	}
 
-	status, _, stderr, err := docker.GetContainerOutput(ctx, etcdHost.DClient, EtcdDownloadBackupContainerName, etcdHost.Address)
+	status, _, stderr, err := docker.GetContainerOutput(ctx, etcdHost.DClient, EtcdDownloadBackupContainerName, etcdHost.Address, true)
 	if status != 0 || err != nil {
 		if removeErr := docker.RemoveContainer(ctx, etcdHost.DClient, etcdHost.Address, EtcdDownloadBackupContainerName); removeErr != nil {
 			log.Warnf(ctx, "Failed to remove container [%s]: %v", removeErr)

services/services.go

@@ -60,13 +60,13 @@ const (
 type RestartFunc func(context.Context, *hosts.Host) error
 
 func runSidekick(ctx context.Context, host *hosts.Host, prsMap map[string]v3.PrivateRegistry, sidecarProcess v3.Process, k8sVersion string) error {
-	isRunning, err := docker.IsContainerRunning(ctx, host.DClient, host.Address, SidekickContainerName, true)
+	exists, err := docker.DoesContainerExist(ctx, host.DClient, host.Address, SidekickContainerName, true)
 	if err != nil {
 		return err
 	}
 	imageCfg, hostCfg, _ := GetProcessConfig(sidecarProcess, host, k8sVersion)
 	isUpgradable := false
-	if isRunning {
+	if exists {
 		isUpgradable, err = docker.IsContainerUpgradable(ctx, host.DClient, imageCfg, hostCfg, SidekickContainerName, host.Address, SidekickServiceName)
 		if err != nil {
 			return err