Added support to perform cluster promotion/demotion

Signed-off-by: Utkarsh Bhatt <[email protected]>

.github/workflows/tests.yml

@@ -709,6 +709,12 @@ jobs:
     - name: Verify RBD mirror
       run : ~/actionutils.sh remote_verify_rbd_mirroring
 
+    - name: Failover site A to Site B
+      run : ~/actionutils.sh remote_failover_to_siteb
+
+    - name: Failback to Site A
+      run : ~/actionutils.sh remote_failback_to_sitea
+
     - name: Disable RBD mirror
       run : ~/actionutils.sh remote_disable_rbd_mirroring
 

docs/how-to/index.rst

@@ -54,6 +54,7 @@ RBD pools and images.
 
    import-remote-cluster
    configure-rbd-mirroring
+   perform-site-failover
 
 Upgrading your cluster
 ----------------------

docs/how-to/perform-site-failover.rst

@@ -0,0 +1,81 @@
+=============================================
+Perform failover for replicated RBD resources
+=============================================
+
+In case of a disaster, all replicated RBD pools can be failed over to a non-primary remote.
+
+An operator can perform promotion on a non-primary cluster, this will inturn promote all replicated rbd
+images in all rbd pools and make them primary. This enables them to be consumed by vms and other workloads.
+
+Prerequisites
+--------------
+1. A primary and a secondary MicroCeph cluster, for example named "primary_cluster" and "secondary_cluster"
+2. primary_cluster has imported configurations from secondary_cluster and vice versa. refer to :doc:`import remote <./import-remote-cluster>`
+3. RBD remote replication is configured for atleast 1 rbd image. refer to :doc:`configure rbd replication <./configure-rbd-mirroring>`
+
+Failover to a non-primary remote cluster
+-----------------------------------------
+List all the resources on 'secondary_cluster' to check primary status.
+
+.. code-block:: none
+
+   sudo microceph remote replication rbd list
+   +-----------+------------+------------+---------------------+
+   | POOL NAME | IMAGE NAME | IS PRIMARY | LAST LOCAL UPDATE   |
+   +-----------+------------+------------+---------------------+
+   | pool_one  | image_one  | false      | 2024-10-14 09:03:17 |
+   | pool_one  | image_two  | false      | 2024-10-14 09:03:17 |
+   +-----------+------------+------------+---------------------+
+
+An operator can perform cluster wide promotion as follows:
+
+.. code-block:: none
+
+   sudo microceph remote replication rbd promote --remote primary_cluster --yes-i-really-mean-it 
+
+Here, <remote> paramter helps microceph filter the resources to promote.
+Since promotion of secondary_cluster may cause a split-brain condition in future,
+it is necessary to pass --yes-i-really-mean-it flag.
+
+Verify RBD remote replication primary status
+---------------------------------------------
+
+List all the resources on 'secondary_cluster' again to check primary status.
+
+.. code-block:: none
+
+   sudo microceph remote replication rbd status pool_one
+   +-----------+------------+------------+---------------------+
+   | POOL NAME | IMAGE NAME | IS PRIMARY | LAST LOCAL UPDATE   |
+   +-----------+------------+------------+---------------------+
+   | pool_one  | image_one  | true       | 2024-10-14 09:06:12 |
+   | pool_one  | image_two  | true       | 2024-10-14 09:06:12 |
+   +-----------+------------+------------+---------------------+
+
+The status shows that there are 2 replicated images and both of them are now primary.
+
+Failback to old primary
+------------------------
+
+Once the disaster struck cluster (primary_cluster) is back online the RBD resources
+can be failed back to it, but, by this time the RBD images at the current primary (secondary_cluster)
+would have diverged from primary_cluster. Thus, to have a clean sync, the operator must decide
+which cluster would be demoted to the non-primary status. This cluster will then receive the 
+RBD mirror updates from the standing primary.
+
+Note: Demotion can cause data loss and hence can only be performed with the 'yes-i-really-mean-it' flag.
+
+At primary_cluster (was primary before disaster), perform demotion.
+.. code-block:: none
+
+   sudo microceph remote replication rbd demote --remote secondary_cluster
+   failed to process demote_replication request for rbd: demotion may cause data loss on this cluster. If you
+   understand the *RISK* and you're *ABSOLUTELY CERTAIN* that is what you want, pass --yes-i-really-mean-it.
+
+Now, again at the 'primary_cluster', perform demotion with --yes-i-really-mean-it flag.
+.. code-block:: none
+
+   sudo microceph remote replication rbd demote --remote secondary_cluster --yes-i-really-mean-it
+
+Note: MicroCeph with demote the primary pools and will issue a resync for all the mirroring images, hence it may
+cause data loss at the old primary cluster.

docs/reference/commands/remote-replication-rbd.rst

@@ -96,3 +96,32 @@ Usage:
 
    --force   forcefully disable replication for rbd resource
 
+``promote``
+------------
+
+Promote local cluster to primary
+
+.. code-block:: none
+
+   microceph remote replication rbd promote [flags]
+
+.. code-block:: none
+
+   --remote         remote MicroCeph cluster name
+   --force          forcefully promote site to primary
+
+``demote``
+------------
+
+Demote local cluster to secondary
+
+Usage:
+
+.. code-block:: none
+
+   microceph remote replication rbd demote [flags]
+
+.. code-block:: none
+
+   --remote         remote MicroCeph cluster name
+

microceph/api/ops_replication.go

@@ -31,6 +31,7 @@ var opsReplicationCmd = rest.Endpoint{
 var opsReplicationWorkloadCmd = rest.Endpoint{
 	Path: "ops/replication/{wl}",
 	Get:  rest.EndpointAction{Handler: getOpsReplicationWorkload, ProxyTarget: false},
+	Put:  rest.EndpointAction{Handler: putOpsReplicationWorkload, ProxyTarget: false},
 }
 
 // CRUD Replication
@@ -47,6 +48,12 @@ func getOpsReplicationWorkload(s state.State, r *http.Request) response.Response
 	return cmdOpsReplication(s, r, types.ListReplicationRequest)
 }
 
+// putOpsReplicationWorkload handles site level (promote/demoteR) operation
+func putOpsReplicationWorkload(s state.State, r *http.Request) response.Response {
+	// either promote or demote (already encoded in request)
+	return cmdOpsReplication(s, r, "")
+}
+
 // getOpsReplicationResource handles status operation for a certain resource.
 func getOpsReplicationResource(s state.State, r *http.Request) response.Response {
 	return cmdOpsReplication(s, r, types.StatusReplicationRequest)
@@ -104,6 +111,9 @@ func cmdOpsReplication(s state.State, r *http.Request, patchRequest types.Replic
 		return response.SmartError(fmt.Errorf(""))
 	}
 
+	// TODO: convert this to debug
+	logger.Infof("REPOPS: %s received for %s: %s", req.GetWorkloadRequestType(), wl, resource)
+
 	return handleReplicationRequest(s, r.Context(), req)
 }
 

microceph/api/types/replication.go

@@ -11,9 +11,13 @@ type ReplicationRequestType string
 const (
 	EnableReplicationRequest    ReplicationRequestType = "POST-" + constants.EventEnableReplication
 	ConfigureReplicationRequest ReplicationRequestType = "PUT-" + constants.EventConfigureReplication
-	DisableReplicationRequest   ReplicationRequestType = "DELETE-" + constants.EventDisableReplication
-	StatusReplicationRequest    ReplicationRequestType = "GET-" + constants.EventStatusReplication
-	ListReplicationRequest      ReplicationRequestType = "GET-" + constants.EventListReplication
+	PromoteReplicationRequest   ReplicationRequestType = "PUT-" + constants.EventPromoteReplication
+	DemoteReplicationRequest    ReplicationRequestType = "PUT-" + constants.EventDemoteReplication
+	// Delete Requests
+	DisableReplicationRequest ReplicationRequestType = "DELETE-" + constants.EventDisableReplication
+	// Get Requests
+	StatusReplicationRequest ReplicationRequestType = "GET-" + constants.EventStatusReplication
+	ListReplicationRequest   ReplicationRequestType = "GET-" + constants.EventListReplication
 )
 
 type CephWorkloadType string

microceph/ceph/rbd_mirror.go

@@ -211,8 +211,8 @@ func DisablePoolMirroring(pool string, peer RbdReplicationPeer, localName string
 	return nil
 }
 
-// DisableMirroringAllImagesInPool disables mirroring for all images for a pool enabled in pool mirroring mode.
-func DisableMirroringAllImagesInPool(poolName string) error {
+// DisableAllMirroringImagesInPool disables mirroring for all images for a pool enabled in pool mirroring mode.
+func DisableAllMirroringImagesInPool(poolName string) error {
 	poolStatus, err := GetRbdMirrorVerbosePoolStatus(poolName, "", "")
 	if err != nil {
 		err := fmt.Errorf("failed to fetch status for %s pool: %v", poolName, err)
@@ -233,6 +233,29 @@ func DisableMirroringAllImagesInPool(poolName string) error {
 	return nil
 }
 
+// ResyncAllMirroringImagesInPool triggers a resync for all mirroring images inside a mirroring pool.
+func ResyncAllMirroringImagesInPool(poolName string) error {
+	poolStatus, err := GetRbdMirrorVerbosePoolStatus(poolName, "", "")
+	if err != nil {
+		err := fmt.Errorf("failed to fetch status for %s pool: %v", poolName, err)
+		logger.Error(err.Error())
+		return err
+	}
+
+	flaggedImages := []string{}
+	for _, image := range poolStatus.Images {
+		err := flagImageForResync(poolName, image.Name)
+		if err != nil {
+			return fmt.Errorf("failed to resync %s/%s", poolName, image.Name)
+		}
+		flaggedImages = append(flaggedImages, image.Name)
+	}
+
+	// TODO: convert to debug
+	logger.Infof("REPRBD: Resynced %v images in %s pool.", flaggedImages, poolName)
+	return nil
+}
+
 // getPeerUUID returns the peer ID for the requested peer name.
 func getPeerUUID(pool string, peerName string, client string, cluster string) string {
 	poolInfo, err := GetRbdMirrorPoolInfo(pool, cluster, client)
@@ -294,6 +317,7 @@ func BootstrapPeer(pool string, localName string, remoteName string) error {
 }
 
 // ############################# Ceph Commands #############################
+// configurePoolMirroring enables/disables mirroring for a pool.
 func configurePoolMirroring(pool string, mode types.RbdResourceType, localName string, remoteName string) error {
 	var args []string
 	if mode == types.RbdResourceDisabled {
@@ -347,6 +371,7 @@ func configureImageMirroring(req types.RbdReplicationRequest) error {
 	return nil
 }
 
+// getSnapshotSchedule fetches the schedule of the snapshots.
 func getSnapshotSchedule(pool string, image string) (imageSnapshotSchedule, error) {
 	if len(pool) == 0 || len(image) == 0 {
 		return imageSnapshotSchedule{}, fmt.Errorf("ImageName(%s/%s) not complete", pool, image)
@@ -464,6 +489,42 @@ func configureImageFeatures(pool string, image string, op string, feature string
 	return nil
 }
 
+// enableImageFeatures enables the list of rbd features on the requested resource.
+func enableRbdImageFeatures(poolName string, imageName string, features []string) error {
+	for _, feature := range features {
+		err := configureImageFeatures(poolName, imageName, "enable", feature)
+		if err != nil && !strings.Contains(err.Error(), "one or more requested features are already enabled") {
+			return err
+		}
+	}
+	return nil
+}
+
+// disableRbdImageFeatures disables the list of rbd features on the requested resource.
+func disableRbdImageFeatures(poolName string, imageName string, features []string) error {
+	for _, feature := range features {
+		err := configureImageFeatures(poolName, imageName, "disable", feature)
+		if err != nil {
+			return err
+		}
+	}
+	return nil
+}
+
+// flagImageForResync flags requested mirroring image in the given pool for resync.
+func flagImageForResync(poolName string, imageName string) error {
+	args := []string{
+		"mirror", "image", "resync", fmt.Sprintf("%s/%s", poolName, imageName),
+	}
+
+	_, err := processExec.RunCommand("rbd", args...)
+	if err != nil {
+		return err
+	}
+
+	return nil
+}
+
 // peerBootstrapCreate generates peer bootstrap token on remote ceph cluster.
 func peerBootstrapCreate(pool string, client string, cluster string) (string, error) {
 	args := []string{
@@ -525,6 +586,46 @@ func peerRemove(pool string, peerId string, localName string, remoteName string)
 	return nil
 }
 
+func promotePool(poolName string, isForce bool, remoteName string, localName string) error {
+	args := []string{
+		"mirror", "pool", "promote", poolName,
+	}
+
+	if isForce {
+		args = append(args, "--force")
+	}
+
+	// add --cluster and --id args
+	args = appendRemoteClusterArgs(args, remoteName, localName)
+
+	output, err := processExec.RunCommand("rbd", args...)
+	if err != nil {
+		return fmt.Errorf("failed to promote pool(%s): %v", poolName, err)
+	}
+
+	// TODO: Change to debugf
+	logger.Infof("REPRBD: Promotion Output: %s", output)
+	return nil
+}
+
+func demotePool(poolName string, remoteName string, localName string) error {
+	args := []string{
+		"mirror", "pool", "demote", poolName,
+	}
+
+	// add --cluster and --id args
+	args = appendRemoteClusterArgs(args, remoteName, localName)
+
+	output, err := processExec.RunCommand("rbd", args...)
+	if err != nil {
+		return fmt.Errorf("failed to promote pool(%s): %v", poolName, err)
+	}
+
+	// TODO: Change to debugf
+	logger.Infof("REPRBD: Demotion Output: %s", output)
+	return nil
+}
+
 // ########################### HELPERS ###########################
 
 func IsRemoteConfiguredForRbdMirror(remoteName string) bool {

microceph/ceph/rbd_mirror_test.go

@@ -1,6 +1,7 @@
 package ceph
 
 import (
+	"fmt"
 	"os"
 	"testing"
 
@@ -93,3 +94,42 @@ func (ks *RbdMirrorSuite) TestPoolInfo() {
 	assert.Equal(ks.T(), resp.LocalSiteName, "magical")
 	assert.Equal(ks.T(), resp.Peers[0].RemoteName, "simple")
 }
+func (ks *RbdMirrorSuite) TestPromotePoolOnSecondary() {
+	r := mocks.NewRunner(ks.T())
+	output, _ := os.ReadFile("./test_assets/rbd_mirror_promote_secondary_failure.txt")
+
+	// mocks and expectations
+	r.On("RunCommand", []interface{}{
+		"rbd", "mirror", "pool", "promote", "pool"}...).Return("", fmt.Errorf("%s", string(output))).Once()
+	r.On("RunCommand", []interface{}{
+		"rbd", "mirror", "pool", "promote", "pool", "--force"}...).Return("ok", nil).Once()
+	processExec = r
+
+	// Test stardard promotion.
+	err := handlePoolPromotion("pool", false)
+	assert.ErrorContains(ks.T(), err, "If you understand the *RISK* and you're *ABSOLUTELY CERTAIN*")
+
+	err = handlePoolPromotion("pool", true)
+	assert.NoError(ks.T(), err)
+}
+
+func (ks *RbdMirrorSuite) TestDemotePoolOnSecondary() {
+	r := mocks.NewRunner(ks.T())
+
+	output, _ := os.ReadFile("./test_assets/rbd_mirror_verbose_pool_status.json")
+
+	// mocks and expectations
+	r.On("RunCommand", []interface{}{
+		"rbd", "mirror", "pool", "demote", "pool"}...).Return("ok", nil).Once()
+	r.On("RunCommand", []interface{}{
+		"rbd", "mirror", "pool", "status", "pool", "--verbose", "--format", "json"}...).Return(string(output), nil).Once()
+	r.On("RunCommand", []interface{}{
+		"rbd", "mirror", "image", "resync", "pool/image_one"}...).Return("ok", nil).Once()
+	r.On("RunCommand", []interface{}{
+		"rbd", "mirror", "image", "resync", "pool/image_two"}...).Return("ok", nil).Once()
+	processExec = r
+
+	// Test stardard promotion.
+	err := handlePoolDemotion("pool")
+	assert.NoError(ks.T(), err)
+}