[
https://issues.apache.org/jira/browse/HBASE-23739?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17141288#comment-17141288
]
Anoop Sam John commented on HBASE-23739:
----------------------------------------
[~zghao] The wal split op might create 1+ HFiles for a region while doing
split for one WAL file right? (depending on how many times the EntryBuffers
needed to be flushed for this region). And like that we might have N WAL files
also.
Can we think of a way such that after the wal split and region open, when we
req a compaction, we will prioritize those files (created by split op) and
compact all of them together into a single file? This is not only related to
this issue discussion but general. Because the split to HFiles might be many
small HFiles and if we have a way to select only those small files and compact,
we can make sure to reduce many small files with min IO effort.
So we will see the issue when we split META wal and request for the FD for meta
table right? Stack what u r trying out is to return the META FD even if Master
init is not yet over?
> BoundedRecoveredHFilesOutputSink should read the table descriptor directly
> --------------------------------------------------------------------------
>
> Key: HBASE-23739
> URL: https://issues.apache.org/jira/browse/HBASE-23739
> Project: HBase
> Issue Type: Sub-task
> Reporter: Guanghao Zhang
> Assignee: Guanghao Zhang
> Priority: Major
> Fix For: 3.0.0-alpha-1, 2.3.0
>
>
> Read from meta or filesystem?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
