[
https://issues.apache.org/jira/browse/HBASE-11610?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14097205#comment-14097205
]
Virag Kothari commented on HBASE-11610:
---------------------------------------
bq. With threadlocal, how is the HTable closed?
Good question. If we were to close it, we would need to store references to it
which is expensive. So, instead if we skip the close it should be fine because
the way htable is created, it wont shutdown the pool or close the connection
(Both cleanupPoolOnClose and cleanupConnectionOnClose is false). So the only
thing it would do is flushCommits but put() will take care of that as
autoflush is true by default.
bq. With more connections, we can achieve some parallelism at the network/RPC
layer.
Yeah, more TCP connections are helping. I have seen the write throughput on
meta region increase to around 450K/min from around 380k/min (~15%) when trying
to assign 1M regions. Tweaking this configuration would help us when we to
scale even further.
> Enhance remote meta updates
> ---------------------------
>
> Key: HBASE-11610
> URL: https://issues.apache.org/jira/browse/HBASE-11610
> Project: HBase
> Issue Type: Improvement
> Reporter: Jimmy Xiang
> Assignee: Virag Kothari
> Attachments: HBASE-11610.patch
>
>
> Currently, if the meta region is on a regionserver instead of the master,
> meta update is synchronized on one HTable instance. We should be able to do
> better.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
