[HBASE-26448] Make sure we do not flush a region too frequently

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 3.0.0-alpha-3
• Hadoop Flags: Reviewed
• Assignee: Duo Zhang
• Reporter: Duo Zhang
• Created: 11/Nov/21 02:08
• Updated: 01/Jan/22 13:38
• Resolved: 15/Nov/21 16:07

Description

In HBASE-26412, once we found there is an error replicating the edits, we will trigger a flush, and in HBASE-26413, we are likely to also trigger a flush when the pending edits are too large.

We should find a way to avoid flushing the region too much.

And a failure of replicating usually means the region is moved or the target region server is crashed, only a flush after the secondary replica is online will be useful, so flushing the region lots of times before the scedondary replica online is just useless.

Issue Links

• links to: GitHub Pull Request #3847