[
https://issues.apache.org/jira/browse/HBASE-1145?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12669261#action_12669261
]
Nitay Joffe commented on HBASE-1145:
------------------------------------
Eventually, the leader election described above is the way to go. Then any
regionservers can assume the master role and there are no special master
servers. That implementation requires a lot of heavy changes, so for right now
we are going to just do a master lock.
Users will spin up multiple masters, and they will race to grab the master lock
in ZooKeeper. Whoever grabs it creates the ephemeral node and writes their
server address to the ZNode. The rest of the masters will sit around waiting
for the ephemeral node to go away (master goes down) and then the race repeats.
Region servers and clients will go to this location in ZooKeeper to get the
master address. This means we can get rid of the hbase.master property in
hbase-site.xml. To retain the option of running in "local" mode, we will add a
"run.distributed" property to the hbase-site.xml which defaults to false. For
backwards compatibility, if users have their hbase.master set to something
other than "local", we can override the run.distributed option and produce a
warning that the "hbase.master" option is going away and they should update
their hbase-site.xml with the new "run.distributed" property.
Thoughts?
> Ensure that there is only 1 Master with Zookeeper
> -------------------------------------------------
>
> Key: HBASE-1145
> URL: https://issues.apache.org/jira/browse/HBASE-1145
> Project: Hadoop HBase
> Issue Type: Sub-task
> Reporter: Jean-Daniel Cryans
> Assignee: Nitay Joffe
>
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
