Sergey Shelukhin created HBASE-7316:
---------------------------------------
Summary: HBase region server retries are not always smart. Make
smarter/replace with request timeouts?
Key: HBASE-7316
URL: https://issues.apache.org/jira/browse/HBASE-7316
Project: HBase
Issue Type: Brainstorming
Components: Client
Reporter: Sergey Shelukhin
Priority: Minor
Example can be seen in HBASE-7250.
Server A has region, region is moved to B, B dies, region starts moving to C.
Client goes to A, A sends him to B, B is dead so he refreshes meta (which has
A), it goes to A, which sends it to B, ...
These retries are unnecessary.
Overall, I wonder if retry policy should be replaced by timeout policy,
preferably configurable on per-request basis by client? That sounds like a
reasonable feature to me, so if the timeout is set to high values retries will
not be exhausted due to situations like this.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
