[ https://issues.apache.org/jira/browse/HBASE-4129?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
stack updated HBASE-4129: ------------------------- Attachment: 3872.txt So, I think the issue is that the HRegionInfo comparator is sorting daughters' before parents so we process daughter first, then parent. Rather than change the HRI comparator, a radical move, I just made a version of it over in CatalogJanitor that will sort parents first. This should make it so I do not have to relax requirement that a daughter region exist before I can remove parent. Testing on cluster now. > hbase-3872 added a warn message 'CatalogJanitor: Daughter regiondir does not > exist' that is triggered though its often legit that daughter is not present > --------------------------------------------------------------------------------------------------------------------------------------------------------- > > Key: HBASE-4129 > URL: https://issues.apache.org/jira/browse/HBASE-4129 > Project: HBase > Issue Type: Bug > Affects Versions: 0.90.4 > Reporter: stack > Assignee: stack > Fix For: 0.90.4 > > Attachments: 3872.txt > > > If a daughter region is split before the catalog janitor runs, we'll see: > {code} > 2011-07-22 16:10:26,398 WARN org.apache.hadoop.hbase.master.CatalogJanitor: > Daughter regiondir does not exist: > hdfs://sv4borg227:10000/hbase/TestTable/a1023b2b00fe44c86bd8ae3633f531fa > {code} > Its legit that the daughter region does not exist in this case (it was just > cleaned up by the catalogjanitor). -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira