[
https://issues.apache.org/jira/browse/HBASE-17407?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15798918#comment-15798918
]
stack commented on HBASE-17407:
-------------------------------
Can I help?
As per [~Apache9], onlyIfGreater seems odd.
Some backgound if it will help:
The map on the regionserver of region+family to lowest sequence id is used in
two places. Locally for figuring when it is safe to let go of a regionservers
WALs and then, remotely as on every heartbeat to the master, we send it our
inmemory map. The master then keeps up a running accounting of the oldest
sequenceid/edit in memstore for each familiy of each region. It comes in handy
at crash time when replaying WALs. If we see an edit for a family that has a
sequenceid that is less than what we have in our inmemory map for a particular
region+family, then we skip replaying the edit since we know it already
persisted (out in an hfile). If lowest sequenceid goes backwards, thats not the
end of the world; we will over-replay. If the sequenceid accounting advances
when it should not, we'll skip the replay of edits that should have been
replayed (dataloss).
Our [~Apache9] knows this area of the code because he added finesse backfilling
accounting by column family (where previous we only did accounting at the
coarser region granularity). Finding bugs in implementation is awkward given
the number of moving pieces and the offset between writing and recovery.
> Correct update of maxFlushedSeqId in HRegion
> --------------------------------------------
>
> Key: HBASE-17407
> URL: https://issues.apache.org/jira/browse/HBASE-17407
> Project: HBase
> Issue Type: Bug
> Reporter: Eshcar Hillel
>
> The attribute maxFlushedSeqId in HRegion is used to track the max sequence id
> in the store files and is reported to HMaster. When flushing only part of the
> memstore content this value might be incorrect and may cause data loss.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
