You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Guanghao Zhang (Jira)" <ji...@apache.org> on 2020/10/28 10:13:00 UTC
[jira] [Updated] (HBASE-25225) Create table very slowly if there
are multi regions
[ https://issues.apache.org/jira/browse/HBASE-25225?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Guanghao Zhang updated HBASE-25225:
-----------------------------------
Description:
Run the same UT TestRegionReplicaFailover on my local PC, mvn clean test -Dtest=TestRegionReplicaFailover, branch-2.2 takes 8 mins but branch-2.3 only needs 2 mins.
I found the problem is related to procedure schedule. See the below log:
2020-10-21 13:52:28,097 INFO [PEWorker-1] procedure2.ProcedureExecutor(1427): Finished pid=296, ppid=45, state=SUCCESS; org.apache.hadoop.hbase.master.assignment.OpenRegionProcedure in 1.6250sec
2020-10-21 13:52:28,538 INFO [PEWorker-3] procedure2.ProcedureExecutor(1427): Finished pid=45, ppid=20, state=SUCCESS; TransitRegionStateProcedure table=testLotsOfRegionRepli2, region=50703895da3cb8c942d3197600d549bc, ASSIGN in 59.4330sec
The real assign procedure only cost 1.6 seconds but the TransitRegionStateProcedure cost 59.4 seconds. The pid=45 procedure was initialized at 2020-10-21 13:51:28,666. It was added to TableQueue at 2020-10-21 13:51:28,789. But took xlock to run at 2020-10-21 13:52:24,761. See the below log:
{color:#ff0000}2020-10-21 13:51:28,789{color} DEBUG [PEWorker-4] procedure.MasterProcedureScheduler(352): Add TableQueue(testLotsOfRegionRepli2, xlock=true (20) sharedLock=0 size=25) to run queue because: pid=45, ppid=20, state=RUNNABLE:REGION_STATE_TRANSITION_GET_ASSIGN_CANDIDATE; TransitRegionStateProcedure table=testLotsOfRegionRepli2, region=50703895da3cb8c942d3197600d549bc, ASSIGN has the excusive lock access
{color:#ff0000}2020-10-21 13:52:24,761{color} INFO [PEWorker-2] procedure.MasterProcedureScheduler(737): Took xlock for pid=45, ppid=20, state=RUNNABLE:REGION_STATE_TRANSITION_GET_ASSIGN_CANDIDATE; TransitRegionStateProcedure table=testLotsOfRegionRepli2, region=50703895da3cb8c942d3197600d549bc, ASSIGN
But when I tried this UT on another PC, it only cost 2 mins, which is the same with branch-2.3. It is weird.
Marked this as blocker for release 2.2.7.
If you are interested for this, please run "mvn clean test -Dtest=TestRegionReplicaFailover" and comment the cost time here. Thanks.
was:
Run the same UT TestRegionReplicaFailover on my local PC, mvn clean test -Dtest=TestRegionReplicaFailover, branch-2.2 takes 8 mins but branch-2.3 only needs 2 mins.
I found the problem is related to procedure schedule. See the below log:
2020-10-21 13:52:28,097 INFO [PEWorker-1] procedure2.ProcedureExecutor(1427): Finished pid=296, ppid=45, state=SUCCESS; org.apache.hadoop.hbase.master.assignment.OpenRegionProcedure in 1.6250sec
2020-10-21 13:52:28,538 INFO [PEWorker-3] procedure2.ProcedureExecutor(1427): Finished pid=45, ppid=20, state=SUCCESS; TransitRegionStateProcedure table=testLotsOfRegionRepli2, region=50703895da3cb8c942d3197600d549bc, ASSIGN in 59.4330sec
The real assign procedure only cost 1.6 seconds but the TransitRegionStateProcedure cost 59.4 seconds. The pid=45 procedure was initialized at 2020-10-21 13:51:28,666. It was added to TableQueue at 2020-10-21 13:51:28,789. But took xlock to run at 2020-10-21 13:52:24,761. See the below log:
{color:#ff0000}2020-10-21 13:51:28,789{color} DEBUG [PEWorker-4] procedure.MasterProcedureScheduler(352): Add TableQueue(testLotsOfRegionRepli2, xlock=true (20) sharedLock=0 size=25) to run queue because: pid=45, ppid=20, state=RUNNABLE:REGION_STATE_TRANSITION_GET_ASSIGN_CANDIDATE; TransitRegionStateProcedure table=testLotsOfRegionRepli2, region=50703895da3cb8c942d3197600d549bc, ASSIGN has the excusive lock access
{color:#ff0000}2020-10-21 13:52:24,761{color} INFO [PEWorker-2] procedure.MasterProcedureScheduler(737): Took xlock for pid=45, ppid=20, state=RUNNABLE:REGION_STATE_TRANSITION_GET_ASSIGN_CANDIDATE; TransitRegionStateProcedure table=testLotsOfRegionRepli2, region=50703895da3cb8c942d3197600d549bc, ASSIGN
But when I tried this UT on another PC, it only cost 2 mins, which is the same with branch-2.3. It is weird.
Marked this as blocker for release 2.2.7.
> Create table very slowly if there are multi regions
> ---------------------------------------------------
>
> Key: HBASE-25225
> URL: https://issues.apache.org/jira/browse/HBASE-25225
> Project: HBase
> Issue Type: Bug
> Affects Versions: 2.2.6
> Reporter: Guanghao Zhang
> Priority: Blocker
>
> Run the same UT TestRegionReplicaFailover on my local PC, mvn clean test -Dtest=TestRegionReplicaFailover, branch-2.2 takes 8 mins but branch-2.3 only needs 2 mins.
>
> I found the problem is related to procedure schedule. See the below log:
> 2020-10-21 13:52:28,097 INFO [PEWorker-1] procedure2.ProcedureExecutor(1427): Finished pid=296, ppid=45, state=SUCCESS; org.apache.hadoop.hbase.master.assignment.OpenRegionProcedure in 1.6250sec
> 2020-10-21 13:52:28,538 INFO [PEWorker-3] procedure2.ProcedureExecutor(1427): Finished pid=45, ppid=20, state=SUCCESS; TransitRegionStateProcedure table=testLotsOfRegionRepli2, region=50703895da3cb8c942d3197600d549bc, ASSIGN in 59.4330sec
>
> The real assign procedure only cost 1.6 seconds but the TransitRegionStateProcedure cost 59.4 seconds. The pid=45 procedure was initialized at 2020-10-21 13:51:28,666. It was added to TableQueue at 2020-10-21 13:51:28,789. But took xlock to run at 2020-10-21 13:52:24,761. See the below log:
> {color:#ff0000}2020-10-21 13:51:28,789{color} DEBUG [PEWorker-4] procedure.MasterProcedureScheduler(352): Add TableQueue(testLotsOfRegionRepli2, xlock=true (20) sharedLock=0 size=25) to run queue because: pid=45, ppid=20, state=RUNNABLE:REGION_STATE_TRANSITION_GET_ASSIGN_CANDIDATE; TransitRegionStateProcedure table=testLotsOfRegionRepli2, region=50703895da3cb8c942d3197600d549bc, ASSIGN has the excusive lock access
> {color:#ff0000}2020-10-21 13:52:24,761{color} INFO [PEWorker-2] procedure.MasterProcedureScheduler(737): Took xlock for pid=45, ppid=20, state=RUNNABLE:REGION_STATE_TRANSITION_GET_ASSIGN_CANDIDATE; TransitRegionStateProcedure table=testLotsOfRegionRepli2, region=50703895da3cb8c942d3197600d549bc, ASSIGN
>
> But when I tried this UT on another PC, it only cost 2 mins, which is the same with branch-2.3. It is weird.
>
> Marked this as blocker for release 2.2.7.
>
> If you are interested for this, please run "mvn clean test -Dtest=TestRegionReplicaFailover" and comment the cost time here. Thanks.
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)