[ https://issues.apache.org/jira/browse/HBASE-13082?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15021607#comment-15021607 ]
stack commented on HBASE-13082: ------------------------------- You think we need to do checkFlushed(); on each seek call? Would doing it on next be enough? Does checkReseek end up calling seek? (Which calls checkFlushed?) Yeah, how often we calling checkFlushed? The less the better. I love the removal of all those locks. Looks beautiful. Is checkFlushed the right name for the method? I see you set flushed to true in 670 flushed = true; in updateReaders ... so should it be checkReadersChanged or checkResetStoreFiles or something? Yeah, why call checkFlushed in shipped? Is that a good place to do it? I'm wondering why we'd call checkFlushed in any place but on the way out of a next call? Why is it a CompactedHFilesCleaner cleaner and not a HFileCleaner? Is it a Cleaner? It closes storefiles. Discharger as in it is discharging the hfile from duty... not it is eligable for archiving... and later delete? Or Acquiter I think this patch is almost there and its a very nice one. > Coarsen StoreScanner locks to RegionScanner > ------------------------------------------- > > Key: HBASE-13082 > URL: https://issues.apache.org/jira/browse/HBASE-13082 > Project: HBase > Issue Type: Bug > Reporter: Lars Hofhansl > Assignee: ramkrishna.s.vasudevan > Attachments: 13082-test.txt, 13082-v2.txt, 13082-v3.txt, > 13082-v4.txt, 13082.txt, 13082.txt, HBASE-13082.pdf, HBASE-13082_1.pdf, > HBASE-13082_12.patch, HBASE-13082_13.patch, HBASE-13082_14.patch, > HBASE-13082_15.patch, HBASE-13082_1_WIP.patch, HBASE-13082_2.pdf, > HBASE-13082_2_WIP.patch, HBASE-13082_3.patch, HBASE-13082_4.patch, > HBASE-13082_9.patch, HBASE-13082_9.patch, HBASE-13082_withoutpatch.jpg, > HBASE-13082_withpatch.jpg, LockVsSynchronized.java, gc.png, gc.png, gc.png, > hits.png, next.png, next.png > > > Continuing where HBASE-10015 left of. > We can avoid locking (and memory fencing) inside StoreScanner by deferring to > the lock already held by the RegionScanner. > In tests this shows quite a scan improvement and reduced CPU (the fences make > the cores wait for memory fetches). > There are some drawbacks too: > * All calls to RegionScanner need to be remain synchronized > * Implementors of coprocessors need to be diligent in following the locking > contract. For example Phoenix does not lock RegionScanner.nextRaw() and > required in the documentation (not picking on Phoenix, this one is my fault > as I told them it's OK) > * possible starving of flushes and compaction with heavy read load. > RegionScanner operations would keep getting the locks and the > flushes/compactions would not be able finalize the set of files. > I'll have a patch soon. -- This message was sent by Atlassian JIRA (v6.3.4#6332)