[
https://issues.apache.org/jira/browse/HBASE-10060?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Lars Hofhansl updated HBASE-10060:
----------------------------------
Attachment: 10060-trunk.txt
Sample trunk patch. Please have a look (not too dissimilar to one of the
attempts on HBASE-10015, but deals with the compactions/flushes as well).
Uses RegionScannerImpl as lock delegate when available, otherwise the called
must hold a lock on the StoreScanner instance.
This notably makes the code a bit more error prone, and it breaks the
abstraction of StoreScanner a bit.
Maybe the lock delegation could be abstracted a bit nicer.
> Unsynchronized scanning
> -----------------------
>
> Key: HBASE-10060
> URL: https://issues.apache.org/jira/browse/HBASE-10060
> Project: HBase
> Issue Type: Bug
> Reporter: Lars Hofhansl
> Attachments: 10060-trunk.txt
>
>
> HBASE-10015 has some lengthy discussion. The solution there ended up
> replacing synchronized with ReentrantLock, which - somewhat surprisingly -
> yielded a non-trivial improvement for tall tables.
> The goal should be to avoid locking in StoreScanner at all. StoreScanner is
> only accessed by a single thread *except* when we have a concurrent flush or
> a compaction, which is rare (we'd acquire and release the lock millions of
> times per second, and compact/flush a few time an hour at the most).
--
This message was sent by Atlassian JIRA
(v6.1#6144)
