[jira] [Created] (HBASE-25015) PerformanceEvaluation with presplit randomWrite test has severe hotspotting
Nick Dimiduk created HBASE-25015:
Summary: PerformanceEvaluation with presplit randomWrite test has
severe hotspotting
Key: HBASE-25015
URL: https://issues.apache.org/jira/browse/HBASE-25015
Project: HBase
Issue Type: Bug
Components: test
Affects Versions: 2.3.1
Reporter: Nick Dimiduk
I'm attempting to apply a load to a small distributed cluster (12 RS). I use
{{--presplit}} and specify 3 regions per region server. {{randomWrite}} test
with {{--nomapred}} and 30 client threads. The result is severe hot-spotting on
a single region region (10's of thousands of reqs/sec) and minimal load (high
10's to low 100's reqs/sec) to the others. It seems the split algorithm and the
load generator do not agree on an even data distribution.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
[jira] [Created] (HBASE-25014) ScheduledChore is never triggered when initalDelay > 1.5*period
Sun Xin created HBASE-25014: --- Summary: ScheduledChore is never triggered when initalDelay > 1.5*period Key: HBASE-25014 URL: https://issues.apache.org/jira/browse/HBASE-25014 Project: HBase Issue Type: Bug Affects Versions: 2.2.5, 2.2.4, 2.2.3, 3.0.0-alpha-1 Reporter: Sun Xin Assignee: Sun Xin Fix For: 3.0.0-alpha-1 In our recent tests, ScheduledChore is never triggered when initalDelay > 1.5*period. The cause of the bug is the following: The trigger time for a ScheduleChore must be within an acceptable time window that is 1.5 * period. see [here|https://github.com/apache/hbase/blob/e5ca9adc54f9f580f85d21d38217afa97aa79d68/hbase-common/src/main/java/org/apache/hadoop/hbase/ScheduledChore.java#L234] timeOfLastRun and timeOfThisRun are two variables that record two adjacent trigger time. [The first initialization of timeOfThisRun|https://github.com/apache/hbase/blob/e5ca9adc54f9f580f85d21d38217afa97aa79d68/hbase-common/src/main/java/org/apache/hadoop/hbase/ScheduledChore.java#L273] is when the ScheduleChore is created, it's not a real trigger time. If we set initialDelay > 1.5 period , after initialDelay, the first time when chore is triggered has exceeded the allowed window. Then [cancel the chore and schedule it again|https://github.com/apache/hbase/blob/e5ca9adc54f9f580f85d21d38217afa97aa79d68/hbase-common/src/main/java/org/apache/hadoop/hbase/ChoreService.java#L176]. So it's stuck in loop when initialDelay > 1.5 period : 1. init timeOfThisRun at a wrong time. 2. wait initalDelay 3. chore trigger, but exceeded the allowed window. 4. cancel chore and schedule it again 5. go step 1. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Created] (HBASE-25013) Improve backup master root cache
Duo Zhang created HBASE-25013: - Summary: Improve backup master root cache Key: HBASE-25013 URL: https://issues.apache.org/jira/browse/HBASE-25013 Project: HBase Issue Type: Sub-task Components: master Reporter: Duo Zhang Assignee: Duo Zhang Could try fan out the edits of root table to backup masters to reduce the latency. And also improve the pull logic to not always reset the cache. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[DISCUSS] Support splittable meta with storing root in master local region(HBASE-24950)
The implementation is almost done, this is the big PR. https://github.com/apache/hbase/pull/2385 There are still some works to be done: 1. Improve the backup master cache 2. Having a clear design on how to support old clients when we split meta 3. Run ITBLL to verify the stability. Want to get some feedbacks on the PR. Thanks.
[jira] [Resolved] (HBASE-24997) [hbase-operator-tools] NPE in RegionsMerger#mergeRegions
[ https://issues.apache.org/jira/browse/HBASE-24997?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Wellington Chevreuil resolved HBASE-24997. -- Resolution: Fixed Thanks for the contribution [~DeanZ]! > [hbase-operator-tools] NPE in RegionsMerger#mergeRegions > > > Key: HBASE-24997 > URL: https://issues.apache.org/jira/browse/HBASE-24997 > Project: HBase > Issue Type: Bug > Components: hbase-operator-tools >Affects Versions: hbase-operator-tools-1.1.0 >Reporter: Baiqiang Zhao >Assignee: Baiqiang Zhao >Priority: Major > > RegionsMerger#mergeRegions will call RegionsMerger#getOpenRegions to get all > OPEN region for target table. When target table's REPLICATION_SCOPE is '1', > RegionsMerger#getOpenRegions may return null. > -- This message was sent by Atlassian Jira (v8.3.4#803005)
