[
https://issues.apache.org/jira/browse/HBASE-21093?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16588828#comment-16588828
]
Duo Zhang commented on HBASE-21093:
-----------------------------------
Checked a successful run and a failing run, there are not much differences on
time before we begin to assign regions.
This is the successful one
{noformat}
2018-08-22 02:22:12,073 DEBUG
[RpcServer.default.FPBQ.Fifo.handler=4,queue=0,port=44370]
procedure2.ProcedureExecutor(1004): Stored pid=75,
state=RUNNABLE:CREATE_TABLE_PRE_OPERATION, hasLock=false; CreateTableProcedure
table=TestMRegions
2018-08-22 02:22:25,003 INFO [RegionOpenAndInitThread-TestMRegions-15]
regionserver.HRegion(1687): Closed
TestMRegions,00000499,1534904531895.e8c30f58fcef29d43e8c5d69b4e582c5.
2018-08-22 02:22:25,543 INFO [PEWorker-1] hbase.MetaTableAccessor(1700):
Updated tableName=TestMRegions, state=ENABLING in hbase:meta
{noformat}
This the failing one
{noformat}
2018-08-21 22:44:38,005 DEBUG
[RpcServer.default.FPBQ.Fifo.handler=3,queue=0,port=48225]
procedure2.ProcedureExecutor(1004): Stored pid=75,
state=RUNNABLE:CREATE_TABLE_PRE_OPERATION, hasLock=false; CreateTableProcedure
table=TestMRegions
2018-08-21 22:44:50,291 INFO [RegionOpenAndInitThread-TestMRegions-13]
regionserver.HRegion(1687): Closed
TestMRegions,00000499,1534891477987.f51d1f6e305eed925b5a4975e84166dd.
2018-08-21 22:44:50,501 INFO [PEWorker-1] hbase.MetaTableAccessor(1700):
Updated tableName=TestMRegions, state=ENABLING in hbase:meta
{noformat}
So I do not think there are much performance difference on the machines? Need
to dig more.
> Increase the dispatch delay for testing DDL procedures
> ------------------------------------------------------
>
> Key: HBASE-21093
> URL: https://issues.apache.org/jira/browse/HBASE-21093
> Project: HBase
> Issue Type: Sub-task
> Components: test
> Reporter: Duo Zhang
> Assignee: Duo Zhang
> Priority: Major
> Attachments: HBASE-21093.patch
>
>
> In TestTableDDLProcedureBase we set the procedure worker number to 1, so on a
> slow machine, we will fail to batch the remote calls to RS with the default
> dispatch delay, and lead to a very long time to finish the bunch of sub TRSPs.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
