[
https://issues.apache.org/jira/browse/HBASE-28950?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
wenhao updated HBASE-28950:
---------------------------
Description:
In my tests, after modifying the 'zookeeper.znode.parent' parameter and
restarting HBase, HBase is unable to complete its initialization with versions
2.3.* and 2.4.* , but this issue does not occur with version 2.2.* .
If the Zookeeper cluster crashes and the data stored in zk cannot be recovered,
and only a new ZK cluster can be set up, the HBase cluster will never be able
to return to a normal state. This problem is particularly severe under certain
failure scenarios.
The following describes the situation after I modified the znode and restarted
HBase. The data in the hbase:meta table was lost, the number of regions
displayed on the Master UI remained at 1, and the meta table could not come
online. Moreover, there were no obvious error messages found in the logs of the
Master or Region Servers.
was:
In my tests, after modifying the 'zookeeper.znode.parent' parameter and
restarting HBase, HBase is unable to complete its initialization with versions
2.3.* and 2.4.*{*}, but this issue does not occur with version 2.2.*{*}.
If the Zookeeper cluster crashes and the data stored in zk cannot be recovered,
and only a new ZK cluster can be set up, the HBase cluster will never be able
to return to a normal state. This problem is particularly severe under certain
failure scenarios.
The following describes the situation after I modified the znode and restarted
HBase. The data in the hbase:meta table was lost, the number of regions
displayed on the Master UI remained at 1, and the meta table could not come
online. Moreover, there were no obvious error messages found in the logs of the
Master or Region Servers.
> After modifying the 'zookeeper.znode.parent' parameter and restarting HBase,
> HBase is unable to complete its initialization.
> ----------------------------------------------------------------------------------------------------------------------------
>
> Key: HBASE-28950
> URL: https://issues.apache.org/jira/browse/HBASE-28950
> Project: HBase
> Issue Type: Bug
> Components: master
> Affects Versions: 2.4.17
> Reporter: wenhao
> Priority: Major
> Attachments: hbase-hbase-master-manager210.bigdata.com.log,
> hbase-hbase-regionserver-manager210.bigdata.com.log,
> hbase-hbase-regionserver-master208.bigdata.com.log,
> hbase-hbase-regionserver-worker209.bigdata.com.log, hbase_meta.png,
> master-ui-home.png, master-ui-procedure.png
>
>
> In my tests, after modifying the 'zookeeper.znode.parent' parameter and
> restarting HBase, HBase is unable to complete its initialization with
> versions 2.3.* and 2.4.* , but this issue does not occur with version 2.2.* .
> If the Zookeeper cluster crashes and the data stored in zk cannot be
> recovered, and only a new ZK cluster can be set up, the HBase cluster will
> never be able to return to a normal state. This problem is particularly
> severe under certain failure scenarios.
>
> The following describes the situation after I modified the znode and
> restarted HBase. The data in the hbase:meta table was lost, the number of
> regions displayed on the Master UI remained at 1, and the meta table could
> not come online. Moreover, there were no obvious error messages found in the
> logs of the Master or Region Servers.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
