support migarte etcd mds and snapshot

Signed-off-by: caoxianfei1 <[email protected]>

cli/command/migrate.go

@@ -39,23 +39,30 @@ import (
 
 var (
 	MIGRATE_ETCD_STEPS = []int{
-		playbook.STOP_SERVICE,
-		playbook.CLEAN_SERVICE, // only container
+		playbook.ADD_ETCD_MEMBER,
 		playbook.PULL_IMAGE,
 		playbook.CREATE_CONTAINER,
 		playbook.SYNC_CONFIG,
+		playbook.AMEND_ETCD_CONFIG,
 		playbook.START_ETCD,
+		playbook.REMOVE_ETCD_MEMBER,
+		playbook.AMEND_SERVER_CONFIG, // modify the etcd endpoint in mds.conf
+		playbook.RESTART_SERVICE,     // restart all mds then modify the etcd endpoint
+		playbook.STOP_SERVICE,
+		playbook.CLEAN_SERVICE, // only container
 		playbook.UPDATE_TOPOLOGY,
 	}
 
 	// mds
 	MIGRATE_MDS_STEPS = []int{
-		playbook.STOP_SERVICE,
-		playbook.CLEAN_SERVICE, // only container
 		playbook.PULL_IMAGE,
 		playbook.CREATE_CONTAINER,
 		playbook.SYNC_CONFIG,
 		playbook.START_MDS,
+		playbook.AMEND_SERVER_CONFIG, // modify the mds.listen.addr in metaserver.conf
+		playbook.RESTART_SERVICE,     // restart all metaserver then modify the mds.listen.addr
+		playbook.STOP_SERVICE,
+		playbook.CLEAN_SERVICE, // only container
 		playbook.UPDATE_TOPOLOGY,
 	}
 
@@ -67,6 +74,8 @@ var (
 		playbook.CREATE_CONTAINER,
 		playbook.SYNC_CONFIG,
 		playbook.START_SNAPSHOTCLONE,
+		playbook.AMEND_SERVER_CONFIG, // modify the mds.listen.addr in metaserver.conf
+		playbook.RESTART_SERVICE,     // restart all metaserver then modify the mds.listen.addr
 		playbook.UPDATE_TOPOLOGY,
 	}
 
@@ -157,7 +166,7 @@ func checkMigrateTopology(curveadm *cli.CurveAdm, data string) error {
 	} else if len(dcs2add) < len(dcs2del) {
 		return errno.ERR_DELETE_SERVICE_WHILE_MIGRATING_IS_DENIED
 	}
-	// len(dcs2add) == len(dcs2del)
+
 	if len(dcs2add) == 0 {
 		return errno.ERR_NO_SERVICES_FOR_MIGRATING
 	}
@@ -199,6 +208,7 @@ func genMigratePlaybook(curveadm *cli.CurveAdm,
 	migrates := getMigrates(curveadm, data)
 	role := migrates[0].From.GetRole()
 	steps := MIGRATE_ROLE_STEPS[role]
+	etcdDCs := curveadm.FilterDeployConfigByRole(dcs, topology.ROLE_ETCD)
 
 	// post clean
 	if options.clean {
@@ -221,10 +231,25 @@ func genMigratePlaybook(curveadm *cli.CurveAdm,
 		config := dcs2add
 		switch step {
 		case playbook.STOP_SERVICE,
-			playbook.CLEAN_SERVICE:
+			playbook.CLEAN_SERVICE,
+			playbook.ADD_ETCD_MEMBER,
+			playbook.REMOVE_ETCD_MEMBER:
 			config = dcs2del
 		case playbook.BACKUP_ETCD_DATA:
 			config = curveadm.FilterDeployConfigByRole(dcs, topology.ROLE_ETCD)
+		// 1. migrate etcd, need to override mds config and restart all mds
+		// 2. (FS)migrate mds, need to override metaserver config and restart all metaservers
+		// 3. (BS)migrate mds, need to override chunkserver and snapshot config and restart all chunkservers and snapshotclones
+		case playbook.AMEND_SERVER_CONFIG,
+			playbook.RESTART_SERVICE:
+			if role == topology.ROLE_ETCD {
+				config = curveadm.FilterDeployConfigByRole(dcs, topology.ROLE_MDS)
+			} else if role == topology.ROLE_MDS && dcs[0].GetKind() == topology.KIND_CURVEFS {
+				config = curveadm.FilterDeployConfigByRole(dcs, topology.ROLE_METASERVER)
+			} else if role == topology.ROLE_MDS && dcs[0].GetKind() == topology.KIND_CURVEBS {
+				config = curveadm.FilterDeployConfigByRole(dcs, topology.ROLE_CHUNKSERVER)
+				config = append(config, curveadm.FilterDeployConfigByRole(dcs, topology.ROLE_SNAPSHOTCLONE)...)
+			}
 		case
 			playbook.CREATE_PHYSICAL_POOL,
 			playbook.CREATE_LOGICAL_POOL,
@@ -251,6 +276,11 @@ func genMigratePlaybook(curveadm *cli.CurveAdm,
 			optionsKV[comm.KEY_POOLSET] = poolset
 		case playbook.UPDATE_TOPOLOGY:
 			optionsKV[comm.KEY_NEW_TOPOLOGY_DATA] = data
+		case playbook.ADD_ETCD_MEMBER,
+			playbook.AMEND_ETCD_CONFIG,
+			playbook.AMEND_SERVER_CONFIG:
+			optionsKV[comm.KEY_MIGRATE_SERVERS] = migrates
+			optionsKV[comm.KEY_CLUSTER_DCS] = etcdDCs
 		}
 
 		pb.AddStep(&playbook.PlaybookStep{

client.yaml

@@ -0,0 +1,16 @@
+kind: curvefs
+s3.ak: curve
+s3.sk: Netease@2023
+s3.endpoint: 10.182.26.46:19005
+s3.bucket_name: curvefs
+container_image: quay.io/opencurve/curve/curvefs:v2.7.0-rc1_d8a1137
+mdsOpt.rpcRetryOpt.addrs: 10.182.26.46:6700,10.182.26.35:6700,10.182.26.36:6700
+log_dir: /home/caoxianfei/client/logs
+data_dir: /mnt/v27cache
+diskCache.fullRatio: 90
+diskCache.safeRatio: 70
+diskCache.trimRatio: 50
+diskCache.maxUsableSpaceBytes: 10737418240
+diskCache.maxFileNums: 1000000
+diskCache.trimCheckIntervalSec: 5
+client.loglevel: 9

internal/common/common.go

@@ -57,6 +57,7 @@ const (
 	// migrate
 	KEY_MIGRATE_STATUS        = "MIGRATE_STATUS"
 	KEY_MIGRATE_COMMON_STATUS = "MIGRATE_COMMON_STATUS"
+	KEY_CLUSTER_DCS           = "CLUSTER_DCS"
 
 	// check
 	KEY_CHECK_WITH_WEAK          = "CHECK_WITH_WEAK"

internal/configure/topology/dc_get.go

@@ -121,7 +121,11 @@ func (dc *DeployConfig) GetInstances() int                   { return dc.instanc
 func (dc *DeployConfig) GetHostSequence() int                { return dc.hostSequence }
 func (dc *DeployConfig) GetInstancesSequence() int           { return dc.instancesSequence }
 func (dc *DeployConfig) GetServiceConfig() map[string]string { return dc.serviceConfig }
-func (dc *DeployConfig) GetVariables() *variable.Variables   { return dc.variables }
+func (dc *DeployConfig) SetServiceConfig(key, value string) {
+	dc.serviceConfig[key] = value
+}
+
+func (dc *DeployConfig) GetVariables() *variable.Variables { return dc.variables }
 
 // (2): config item
 func (dc *DeployConfig) GetPrefix() string           { return dc.getString(CONFIG_PREFIX) }

internal/configure/topology/variables.go

@@ -118,9 +118,9 @@ var (
 		{name: "cluster_mds_dummy_addr"},
 		{name: "cluster_mds_dummy_port"},
 		{name: "cluster_chunkserver_addr", kind: []string{KIND_CURVEBS}},
-		{name: "cluster_snapshotclone_addr", kind: []string{KIND_CURVEBS}},
+		{name: "cluster_snapshotclone_addr"},
 		{name: "cluster_snapshotclone_proxy_addr", kind: []string{KIND_CURVEBS}},
-		{name: "cluster_snapshotclone_dummy_port", kind: []string{KIND_CURVEBS}},
+		{name: "cluster_snapshotclone_dummy_port"},
 		{name: "cluster_snapshotclone_nginx_upstream", kind: []string{KIND_CURVEBS}},
 		{name: "cluster_snapshot_addr"},       // tools-v2: compatible with some old version image
 		{name: "cluster_snapshot_dummy_addr"}, // tools-v2

internal/errno/errno.go

@@ -404,6 +404,9 @@ var (
 	ERR_GET_CHUNKSERVER_COPYSET              = EC(410026, "failed to get chunkserver copyset")
 	ERR_GET_MIGRATE_COPYSET                  = EC(410027, "migrate chunkserver copyset info must be 2")
 	ERR_CONTAINER_NOT_REMOVED                = EC(410027, "container not removed")
+	ERR_GET_CLUSTER_ETCD_ADDR                = EC(410028, "failed to get cluster_etcd_addr variable")
+	ERR_ADD_ETCD_MEMEBER                     = EC(410029, "failed to add etcd member to existing etcd cluster")
+	ERR_REMOVE_ETCD_MEMBER                   = EC(410030, "failed to remove etcd member from existing etcd cluster")
 	// 420: common (curvebs client)
 	ERR_VOLUME_ALREADY_MAPPED             = EC(420000, "volume already mapped")
 	ERR_VOLUME_CONTAINER_LOSED            = EC(420001, "volume container is losed")

internal/playbook/factory.go

@@ -84,6 +84,10 @@ const (
 	INSTALL_CLIENT
 	UNINSTALL_CLIENT
 	ATTACH_LEADER_OR_RANDOM_CONTAINER
+	ADD_ETCD_MEMBER
+	AMEND_ETCD_CONFIG
+	AMEND_SERVER_CONFIG
+	REMOVE_ETCD_MEMBER
 
 	// bs
 	FORMAT_CHUNKFILE_POOL
@@ -251,6 +255,14 @@ func (p *Playbook) createTasks(step *PlaybookStep) (*tasks.Tasks, error) {
 			t, err = comm.NewInstallClientTask(curveadm, config.GetCC(i))
 		case UNINSTALL_CLIENT:
 			t, err = comm.NewUninstallClientTask(curveadm, nil)
+		case ADD_ETCD_MEMBER:
+			t, err = comm.NewAddEtcdMemberTask(curveadm, config.GetDC(i))
+		case AMEND_ETCD_CONFIG:
+			t, err = comm.NewAmendEtcdConfigTask(curveadm, config.GetDC(i))
+		case AMEND_SERVER_CONFIG:
+			t, err = comm.NewAmendServerConfigTask(curveadm, config.GetDC(i))
+		case REMOVE_ETCD_MEMBER:
+			t, err = comm.NewRemoveEtcdMemberTask(curveadm, config.GetDC(i))
 		// bs
 		case FORMAT_CHUNKFILE_POOL:
 			t, err = bs.NewFormatChunkfilePoolTask(curveadm, config.GetFC(i))

internal/task/scripts/enable_etcd_auth.go

@@ -12,13 +12,13 @@
 *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 *  See the License for the specific language governing permissions and
 *  limitations under the License.
-*/
+ */
 
 /*
 * Project: Curveadm
 * Created Date: 2023-08-02
 * Author: wanghai (SeanHai)
-*/
+ */
 
 package scripts
 

internal/task/scripts/script.go

@@ -39,6 +39,10 @@ var (
 	WAIT string
 	//go:embed shell/report.sh
 	REPORT string
+	//go:embed shell/add_etcd.sh
+	ADD_ETCD string
+	//go:embed shell/remove_etcd.sh
+	REMOVE_ETCD string
 
 	// CurveBS
 

internal/task/scripts/shell/add_etcd.sh

@@ -0,0 +1,47 @@
+#!/usr/bin/env bash
+
+# Usage: 
+# Example: 
+# Created Date: 2023-12-15
+# Author: Caoxianfei
+
+etcdctl=$1
+endpoints=$2
+old_name=$3
+new_name=$4
+new_peer_url=$5
+
+tmplog=/tmp/_curveadm_add_etcd_
+
+output=$(${etcdctl} --endpoints=${endpoints} member list)
+if [ $? -ne 0 ]; then 
+    echo "failed to list all etcd members"
+    exit 1
+fi
+
+# if member has added, then skip
+id=$(echo "$output" | awk -v name="$old_name" -F ', ' '$3 == name {print $1}')
+if [ -z "${id}" ]; then
+    echo "EXIST"
+    exit 0
+fi
+
+${etcdctl} --endpoints=${endpoints} member add ${new_name} --peer-urls ${new_peer_url} > ${tmplog} 2>&1
+if [ $? -ne 0 ]; then
+    if cat ${tmplog} | grep -q "Peer URLs already exists"; then 
+        exit 0
+    else
+        exit 1
+    fi
+fi
+
+
+# ${etcdctl} --endpoints=${endpoints} member remove ${id}
+# if [ $? -ne 0 ]; then
+#     echo "failed to remove member ${old_name}"
+#     exit 1
+# fi
+
+
+
+

internal/task/scripts/shell/remove_etcd.sh

@@ -0,0 +1,33 @@
+#!/usr/bin/env bash
+
+# Usage: 
+# Example: 
+# Created Date: 2023-12-15
+# Author: Caoxianfei
+
+etcdctl=$1
+endpoints=$2
+old_name=$3
+
+output=$(${etcdctl} --endpoints=${endpoints} member list)
+if [ $? -ne 0 ]; then 
+    echo "failed to list all etcd members"
+    exit 1
+fi
+
+id=$(echo "$output" | awk -v name="$old_name" -F ', ' '$3 == name {print $1}')
+# if not found the name then exit 0
+if [ -z "${id}" ]; then
+    echo "NOTEXIST"
+    exit 0
+fi
+
+${etcdctl} --endpoints=${endpoints} member remove ${id}
+if [ $? -ne 0 ]; then
+    echo "failed to remove member ${old_name}"
+    exit 1
+fi
+
+
+
+