[
https://issues.apache.org/jira/browse/HBASE-4442?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ming Ma updated HBASE-4442:
---------------------------
Description:
We use a script that restart RS on a rolling fashion and found couple new
issues in HBase in the area of server shutdown and region assignments. Running
such test on a cluster will help to verify changes made in the area of server
shutdown and regions assignment. The script could do the followings:
1. Based on the cluster config file, get the list of RSs and master server.
2. Restart RSs and Master periodically. The parameters like the sleep time
between stop and start, the percentage of machines down at at given time, etc.
should be configurable.
3. Use hbck or other ways to verify the cluster is moving. There will be
inconsistency at a given time when you have lots of regions in transition. So
that is ok. Script should mark test failure only when things aren't moving, for
example, .META. isn't online for > timeoutmonitor period.
was:
We use a script that restart RS on a rolling fashion and found couple new
issues in HBase in the area of server shutdown and region assignments. Running
such test on a cluster will help to verify changes made in the area of server
shutdown and regions assignment. The script could do the followings:
1. Based on the cluster config file, get the list of RSs and master server.
2. Restart RSs and Master periodically. The parameters like the sleep time
between stop and start, the percentage of machines down at at given time, etc.
should be configurable.
3. Use hbck or other ways to verify the cluster is moving. There will be
inconsistency at a given time when you have lots of regions in transition. So
that is ok. Script should mark test failure only when things are moving, for
example, .META. isn't online for > timeoutmonitor period.
good point, Andy.
> Provide test script to do rolling restart of RSs, master and do verification
> ----------------------------------------------------------------------------
>
> Key: HBASE-4442
> URL: https://issues.apache.org/jira/browse/HBASE-4442
> Project: HBase
> Issue Type: Improvement
> Components: test
> Reporter: Ming Ma
> Assignee: Michael Weng
>
> We use a script that restart RS on a rolling fashion and found couple new
> issues in HBase in the area of server shutdown and region assignments.
> Running such test on a cluster will help to verify changes made in the area
> of server shutdown and regions assignment. The script could do the followings:
> 1. Based on the cluster config file, get the list of RSs and master server.
> 2. Restart RSs and Master periodically. The parameters like the sleep time
> between stop and start, the percentage of machines down at at given time,
> etc. should be configurable.
> 3. Use hbck or other ways to verify the cluster is moving. There will be
> inconsistency at a given time when you have lots of regions in transition. So
> that is ok. Script should mark test failure only when things aren't moving,
> for example, .META. isn't online for > timeoutmonitor period.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
