[
https://issues.apache.org/jira/browse/HBASE-17407?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15794617#comment-15794617
]
Eshcar Hillel commented on HBASE-17407:
---------------------------------------
In HRegion::internalFlushCacheAndCommit() we have the following piece of code
{code}
// If we get to here, the HStores have been written.
for(Store storeToFlush :storesToFlush) {
((HStore) storeToFlush).finalizeFlush();
}
if (wal != null) {
wal.completeCacheFlush(this.getRegionInfo().getEncodedNameAsBytes());
}
// Record latest flush time
for (Store store: storesToFlush) {
this.lastStoreFlushTimeMap.put(store, startTime);
}
this.maxFlushedSeqId = flushedSeqId;
this.lastFlushOpSeqId = flushOpSeqId;
{code}
The method finalizeFlush() was added when we added the compacting memstore
exactly for the purpose of updating the WAL with the correct sequence number
before the flush operation is completed.
Indeed, as the code indicates maxFlushSeqId is set with the parameter passed to
the method and is not re-computed based on the current state of the stores.
2 options to fix this
(1) see if we can pass the correct sequence id to the method (maybe we do??)
(2) build upon the finalizeFlush method which finds lowest un-flushed
SequenceId per store. We can return the *lowest unflushed seq id* per store,
find the minimum value and then derive the *maximal flushed seq id*.
> 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)
