[
https://issues.apache.org/jira/browse/HBASE-24580?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17138771#comment-17138771
]
Nick Dimiduk commented on HBASE-24580:
--------------------------------------
I don't actually know if this is a good idea or not, but seems like we could
stream these entries to a pbuf file, maybe even reuse our existing WAL format.
We have everything we need for durable streaming, and these entries shouldn't
happen a too terribly high of a clip, so maybe? We'd need another chore to
truncate the wal periodically, is that solved already too?
> Persist map of regions-to-last-flush-sequenceid across restarts
> ---------------------------------------------------------------
>
> Key: HBASE-24580
> URL: https://issues.apache.org/jira/browse/HBASE-24580
> Project: HBase
> Issue Type: Improvement
> Reporter: Michael Stack
> Priority: Minor
>
> A nice suggestion by our [~ndimiduk] was that we persist the
> last-successful-flush-sequenceid per region kept up by an in-memory map in
> Master. We could add it to our nice new general local master store. On
> restart, could rehydrate the map. Would be helpful on cold start in
> particular, especially if a big backlog of WALs to replay.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
