-
Notifications
You must be signed in to change notification settings - Fork 8
The steps to Rolling Restart Hosts
When you run cmux rolling-restart-hosts
, you will always the folling steps.
-
Choose Cloudera Manager to rolling restart roles (only 1).
Choose Cloudera Manager: Cloudera Manager Description - ---------------- ------------- >1 cm1 Test_Cluster1 2 cm2 Test_Cluster2
-
Choose Cluster to rolling restart cluster (only 1).
ROLLING RESTART HOSTS Choose Cluster: Cloudera Manager Cluster Cluster Display Name CDH Ver Secured Hosts ----------------- -------- -------------------- ------- ------- ----- cm1 cluster2 cluster2 5.8.3 Y 9 > cm1 cluster3 cluster3 5.8.2 N 6 cm1 cluster4 cluster4 5.8.3 Y 6 cm1 cluster5 cluster5 5.7.0 N 5 cm1 cluster6 cluster6 5.7.0 N 6 cm1 cluster cluster1 5.3.8 N 1 cm2 - - - N 3 cm2 cluster cluster1 5.4.9 Y 8
-
Choose the ROLE TYPE to rolling restart host(s).
ROLLING RESTART HOSTS * Cloudera Manager : cm1 * Cluster : [cluster3] cluster3 Choose Hosts(s): Hostname Role types ------------------- ----------------------------- ... host003.kakao.cmux ZK(F),RM(A),FC,JN,NN(S),HM(S) host004.kakao.cmux ZK(L),JHS,JN,FC,NN(A),HM(A) > host005.kakao.cmux ZK(F),RM(S),BL,JN,HM(S) host006.kakao.cmux NM,DN,HBTS,RS host007.kakao.cmux NM,DN,HBTS,RS host008.kakao.cmux NM,DN,RS,HBTS > host009.kakao.cmux NM,DN,RS,HBTS host010.kakao.cmux NM,DN,RS,HBTS ...
-
Review your selection.
ROLLING RESTART ROLES --------------------------------------------------------------------------------- * Cloudera Manager : cm1 * Cluster : [cluster3] cluster3 CDH Version : 5.4.9 Secured : Y Zookeeper : zk1.kakao.cmux hbase-tool : hbase-manager-0.98-1.4.1.jar * Hosts host005.kakao.cmux ├── SERVER REPLICATED_FOLLOWER ├── RESOURCEMANAGER STANDBY ├── BALANCER - ├── JOURNALNODE - └── MASTER STANDBY host009.kakao.cmux ├── HBASETHRIFTSERVER - ├── NODEMANAGER - ├── REGIONSERVER - └── DATANODE - ---------------------------------------------------------------------------------
If REGIONSERVER is running on these hosts, display additional information for zookeeper leader and
hbase-manager
. -
Confirm your selection.
Are you sure you want to ROLLING RESTART on the above roles (y|n)? y
-
Set seconds to sleep between each batch.
Set SECONDS TO SLEEP between batches (>= 0 secs): 3
-
Set the max wait time after executing a roll RESTART command.
Set the MAX WAIT TIME after executing the RESTART command (>= 180 secs): 180
-
Choose whether or not you run interactive mode.
Do you want to proceed with INTERACTIVE MODE (y|n)? y
-
If REGIONSERVERs are running on these hosts:
Turn off auto balancer and export assignment of all resgions on this HBase using
hbase-manager
.Disableig auto balancer └── java -jar .../hbase-manager-1.2-1.4.1.jar assign zk1.kakao.cmux balancer off | tail -1 Automatic balancer is turned off. Export assignment of all Regions to /tmp/cm1_cluster3_20170222_130129.exp └── java -jar .../hbase-manager-1.2-1.4.1.jar assign zk1.kakao.cmux export /tmp/cm1_cluster3_20170222_130129.exp Exported
-
Perform rolling restart roles for selected hosts.
Repeat following steps for hosts.
(1) Stop roles
-
Enter a role level maintenance mode.
-
Stop role.
Stop [host005.kakao.cmux] yarn3-NODEMANAGER-9fbfadaba0a58469c2f1691d312d3d60 Enter maintenance mode └── Maintenance owners: ["ROLE"] Stop yarn3-NODEMANAGER-9fbfadaba0a58469c2f1691d312d3d60 Stopped ... Stop [host005.kakao.cmux] hdfs3-DATANODE-9fbfadaba0a58469c2f1691d312d3d60 Enter maintenance mode └── Maintenance owners: ["ROLE"] Stop hdfs3-DATANODE-9fbfadaba0a58469c2f1691d312d3d60 Stopped
If REGIONSERVERs are running on these hosts:
Before stop REGIONSERVER role, move all Regions on this RegionServer to others using
hbase-manager
.If Nameservices for HDFS HA are running on these hosts:
Before stop NAMENODE role, perform failover.
(2) Start roles
-
Start role.
-
Exit maintenance mode.
Start [host005.kakao.cmux] hdfs3-DATANODE-9fbfadaba0a58469c2f1691d312d3d60 Start hdfs3-DATANODE-9fbfadaba0a58469c2f1691d312d3d60 Started Exit maintenance mode └── Maintenance owners: [] ... Start [host005.kakao.cmux] yarn3-NODEMANAGER-9fbfadaba0a58469c2f1691d312d3d60 Start yarn3-NODEMANAGER-9fbfadaba0a58469c2f1691d312d3d60 Started Exit maintenance mode └── Maintenance owners: []
If REGIONSERVERs are running on these hosts:
After start REGIONSERVER role, restore Regions on this role using
hbase-manager
.If JOURNALNODEs are running on this host:
After start JOURNALNODE role, roll the edits of an HDFS Nameservice.
-
-
If you have HBase service
Turn on HBase auto balancer using
hbase-manager
Do you want to turn on auto balancer (y|n)? y Enabling auto Balancer └── java -jar .../hbase-manager-1.2-1.4.1.jar assign zk1.kakao.cmux balancer on | tail -1 Automatic balancer is turned on Finish