Merge pull request #544 from red-hat-storage/sync_ds--release-1.13

Syncing latest changes from release-1.13 for rook

cmd/rook/ceph/osd.go

@@ -326,12 +326,8 @@ func removeOSDs(cmd *cobra.Command, args []string) error {
 		return errors.Wrapf(err, "failed to parse --preserve-pvc flag")
 	}
 
-	exitIfNotSafe := false
-	forceRemovalCallback := func(x int) (bool, bool) {
-		return forceOSDRemovalBool, exitIfNotSafe
-	}
 	// Run OSD remove sequence
-	err = osddaemon.RemoveOSDs(context, &clusterInfo, strings.Split(osdIDsToRemove, ","), preservePVCBool, forceRemovalCallback)
+	err = osddaemon.RemoveOSDs(context, &clusterInfo, strings.Split(osdIDsToRemove, ","), preservePVCBool, forceOSDRemovalBool)
 	if err != nil {
 		rook.TerminateFatal(err)
 	}

deploy/charts/rook-ceph/templates/clusterrole.yaml

@@ -76,7 +76,6 @@ rules:
   # Node access is needed for determining nodes where mons should run
   - nodes
   - nodes/proxy
-  - services
   # Rook watches secrets which it uses to configure access to external resources.
   # e.g., external Ceph cluster or object store
   - secrets
@@ -96,6 +95,7 @@ rules:
   - persistentvolumeclaims
   # Rook creates endpoints for mgr and object store access
   - endpoints
+  - services
   verbs:
   - get
   - list

deploy/examples/common.yaml

@@ -229,7 +229,6 @@ rules:
       # Node access is needed for determining nodes where mons should run
       - nodes
       - nodes/proxy
-      - services
       # Rook watches secrets which it uses to configure access to external resources.
       # e.g., external Ceph cluster or object store
       - secrets
@@ -249,6 +248,7 @@ rules:
       - persistentvolumeclaims
       # Rook creates endpoints for mgr and object store access
       - endpoints
+      - services
     verbs:
       - get
       - list

pkg/daemon/ceph/osd/remove.go

@@ -33,7 +33,7 @@ import (
 )
 
 // RemoveOSDs purges a list of OSDs from the cluster
-func RemoveOSDs(context *clusterd.Context, clusterInfo *client.ClusterInfo, osdsToRemove []string, preservePVC bool, forceRemovalCallback func(osdID int) (bool, bool)) error {
+func RemoveOSDs(context *clusterd.Context, clusterInfo *client.ClusterInfo, osdsToRemove []string, preservePVC, forceOSDRemoval bool) error {
 	// Generate the ceph config for running ceph commands similar to the operator
 	if err := client.WriteCephConfig(context, clusterInfo); err != nil {
 		return errors.Wrap(err, "failed to write the ceph config")
@@ -63,13 +63,13 @@ func RemoveOSDs(context *clusterd.Context, clusterInfo *client.ClusterInfo, osds
 			logger.Infof("osd.%d is marked 'DOWN'", osdID)
 		}
 
-		removeOSD(context, clusterInfo, osdID, preservePVC, forceRemovalCallback)
+		removeOSD(context, clusterInfo, osdID, preservePVC, forceOSDRemoval)
 	}
 
 	return nil
 }
 
-func removeOSD(clusterdContext *clusterd.Context, clusterInfo *client.ClusterInfo, osdID int, preservePVC bool, forceRemovalCallback func(osdID int) (bool, bool)) {
+func removeOSD(clusterdContext *clusterd.Context, clusterInfo *client.ClusterInfo, osdID int, preservePVC, forceOSDRemoval bool) {
 	// Get the host where the OSD is found
 	hostName, err := client.GetCrushHostName(clusterdContext, clusterInfo, osdID)
 	if err != nil {
@@ -83,21 +83,17 @@ func removeOSD(clusterdContext *clusterd.Context, clusterInfo *client.ClusterInf
 	if err != nil {
 		logger.Errorf("failed to exclude osd.%d out of the crush map. %v", osdID, err)
 	}
-	forceRemoval, exitIfNotSafe := forceRemovalCallback(osdID)
+
 	// Check we can remove the OSD
 	// Loop forever until the osd is safe-to-destroy
 	for {
 		isSafeToDestroy, err := client.OsdSafeToDestroy(clusterdContext, clusterInfo, osdID)
 		if err != nil {
 			// If we want to force remove the OSD and there was an error let's break outside of
 			// the loop and proceed with the OSD removal
-
-			if forceRemoval {
+			if forceOSDRemoval {
 				logger.Errorf("failed to check if osd %d is safe to destroy, but force removal is enabled so proceeding with removal. %v", osdID, err)
 				break
-			} else if exitIfNotSafe {
-				logger.Error("osd.%d is not safe to destroy")
-				return
 			} else {
 				logger.Errorf("failed to check if osd %d is safe to destroy, retrying in 1m. %v", osdID, err)
 				time.Sleep(1 * time.Minute)
@@ -111,12 +107,9 @@ func removeOSD(clusterdContext *clusterd.Context, clusterInfo *client.ClusterInf
 			break
 		} else {
 			// If we arrive here and forceOSDRemoval is true, we should proceed with the OSD removal
-			if forceRemoval {
+			if forceOSDRemoval {
 				logger.Infof("osd.%d is NOT ok to destroy but force removal is enabled so proceeding with removal", osdID)
 				break
-			} else if exitIfNotSafe {
-				logger.Error("osd.%d is not safe to destroy")
-				return
 			}
 			// Else we wait until the OSD can be removed
 			logger.Warningf("osd.%d is NOT ok to destroy, retrying in 1m until success", osdID)