[HBASE-6713] Stopping META/ROOT RS may take 50mins when some region is splitting

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.94.1
• Fix Version/s: 0.94.2
• Component/s: regionserver
• Hadoop Flags: Reviewed
• Assignee: Chunhui Shen
• Reporter: Chunhui Shen
• Created: 03/Sep/12 05:55
• Updated: 07/Apr/13 04:56
• Resolved: 10/Sep/12 21:56

Description

When we stop the RS carrying ROOT/META, if it is in the splitting for some region, the whole stopping process may take 50 mins.
The reason is :
1.ROOT/META region is closed when stopping the regionserver
2.The Split Transaction failed updating META and it will retry
3.The retry num is 100, and the total time is about 50 mins as default;
This configuration is set by HConnectionManager#setServerSideHConnectionRetries

I think 50 mins is too long to acceptable, my suggested solution is closing MetaTable regions after the compact/split thread is closed

Issue Links

• is required by: HBASE-5843 Improve HBase MTTR - Mean Time To Recover