[ https://issues.apache.org/jira/browse/HBASE-21035?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16608185#comment-16608185 ]
stack commented on HBASE-21035: ------------------------------- I'm back. My master is aborting after spending fours reconstructing the assignment state from the reading of 300+ WALs. Master then becomes active. In background there are procedures running and finishing... mostly SCPs. Then my master is dying with: 2018-09-07 22:21:58,968 ERROR org.apache.hadoop.hbase.master.HMaster: ***** ABORTING master vc0207.halxg.cloudera.com,22001,1536380265734: Unhandled exception. Starting shutdown. ***** org.apache.hadoop.hbase.client.RetriesExhaustedException: Failed after attempts=31, exceptions: Fri Sep 07 22:21:58 PDT 2018, null, java.net.SocketTimeoutException: callTimeout=60000, callDuration=69864: org.apache.hadoop.hbase.NotServingRegionException: hbase:meta,,1 is not online on vd0412.halxg.cloudera.com,22101,1536380043533 i.e. meta is not online even though the above server's SCP has completed. This is a dirty install so what was up in zk for meta location could be long stale.... but here we have a state where no SCP's running and meta is not online. I need to write the tool to insert a meta assign. It just takes 4 or 5 hours before I know if it is the fix for this problem. And then there is the scan of the hbase:namespace table next. Thinking of waiting on all SCPs to finish before we do our first meta scan .... and if meta is still not online, then, auto-schedule the restore meta procedure ... splitting meta logs inline and then assigning meta. Would this violate your principal [~Apache9]? In other words, I need to write the restore meta procedure -- it would split meta logs and then do the assign of meta -- but I think we should auto-schedule it in the case above. > Meta Table should be able to online even if all procedures are lost > ------------------------------------------------------------------- > > Key: HBASE-21035 > URL: https://issues.apache.org/jira/browse/HBASE-21035 > Project: HBase > Issue Type: Sub-task > Affects Versions: 2.1.0 > Reporter: Allan Yang > Assignee: Allan Yang > Priority: Major > Attachments: HBASE-21035.branch-2.0.001.patch > > > After HBASE-20708, we changed the way we init after master starts. It will > only check WAL dirs and compare to Zookeeper RS nodes to decide which server > need to expire. For servers which's dir is ending with 'SPLITTING', we assure > that there will be a SCP for it. > But, if the server with the meta region crashed before master restarts, and > if all the procedure wals are lost (due to bug, or deleted manually, > whatever), the new restarted master will be stuck when initing. Since no one > will bring meta region online. > Although it is an anomaly case, but I think no matter what happens, we need > to online meta region. Otherwise, we are sitting ducks, noting can be done. -- This message was sent by Atlassian JIRA (v7.6.3#76005)