[
https://issues.apache.org/jira/browse/HBASE-24822?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17185592#comment-17185592
]
Michael Stack commented on HBASE-24822:
---------------------------------------
[~filtertip] Looking at the command help, it says:
{code:java}
Archive the earliest log files. If too many regions need to be flushed,
it will cost a few time.{code}
So, why would I or an operator want to archive the earliest WAL (not 'log')
files? Why would I not just let the system work and remove oldest files
'naturall'y by flushing the Region+Store with the oldest edit when we reach the
WAL limit?
How many will it archive? Is it the internal calculation? Just the oldest so
usually just the oldest WAL?
Do we need a dedicated command? Could it not be an option on the 'flush'
command? The flush command has this option:
{code:java}
hbase> flush 'REGION_SERVER_NAME'{code}
... which will flush all Regions. Could we add an option something like:
{code:java}
flush 'REGION_SERVER_NAME', 'EARLIEST'
{code}
.. which would flush the store w/ earliest edit or region w/ earliest edit?
Thanks.
> Add a command to support archive the earliest log file manually
> ---------------------------------------------------------------
>
> Key: HBASE-24822
> URL: https://issues.apache.org/jira/browse/HBASE-24822
> Project: HBase
> Issue Type: New Feature
> Components: wal
> Reporter: Zheng Wang
> Assignee: Zheng Wang
> Priority: Major
>
> This new command named wal_archive, belong to the group of "tools".
> It is a sync call, if the data related to earliest log file not all
> persisted, they will be flushed immediately.
> Usage:
> {code:java}
> wal_archive 'desktop-ud83s84,16020,1596595277350'{code}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
