James Taylor created PHOENIX-3875:
-------------------------------------
Summary: Ensure that data writes for tables with indexes work when
split occurs
Key: PHOENIX-3875
URL: https://issues.apache.org/jira/browse/PHOENIX-3875
Project: Phoenix
Issue Type: Bug
Reporter: James Taylor
We most often make an initial RPC to send the index meta data to the regions
involved in a batched mutation and then cache the meta data for use during the
processing of update. If a split occurs while the batch mutation is being
processed and the new region ends up on a different region server, we have a
built-in retry in MutationState. I'm not sure how well this scenario is tested.
For example, what happens if the split occurs when the operation is being
handled on the server-side (i.e. like a DELETE or UPSERT SELECT with auto
commit set to true)?
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
