[ https://issues.apache.org/jira/browse/HBASE-2611?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13402775#comment-13402775 ]
Himanshu Vashishtha commented on HBASE-2611: -------------------------------------------- Thanks for the review Ted. I will upload a modified version on the rb. My initial idea of putting it here was to get some feedback on the approach. Yes, it is zk intensive as all other regionservers are competing to do the transaction. But, as soon as one is successful (the first one who create the list and issues the multi command), other regionservers which haven't had a chance to do a listChildern call on the dead regionserver znode will not see anything; and for other regionservers which have created the Ops, the very first Op will fail as the znode has already moved. Zookeeper#multi op is fail fast, it rolls back the transaction on first failure without retrying remaining Ops. I tested it on a 3 RS cluster with average load being 12-14 logs, and it usually is done within seconds after the regionserver failure is noticed. What sort of experiments you are thinking about. On an another note, TestReplication passes. > Handle RS that fails while processing the failure of another one > ---------------------------------------------------------------- > > Key: HBASE-2611 > URL: https://issues.apache.org/jira/browse/HBASE-2611 > Project: HBase > Issue Type: Sub-task > Components: replication > Reporter: Jean-Daniel Cryans > Assignee: Jean-Daniel Cryans > Attachments: HBase-2611-upstream-v1.patch > > > HBASE-2223 doesn't manage region servers that fail while doing the transfer > of HLogs queues from other region servers that failed. Devise a reliable way > to do it. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa For more information on JIRA, see: http://www.atlassian.com/software/jira