Details
-
Improvement
-
Status: Closed
-
Major
-
Resolution: Cannot Reproduce
-
None
-
None
-
None
-
None
Description
I witnessed a case last night where a region open got stuck – it was an idx hbase region running on 0.20.5, seems like this combo doesn't work – and it prevented our opening other regions behind it. We never recovered. Both the table that had the stuck region and all tables that had regions that were in the queue behind this stuck region were now borked. There was no way out.
Add a timeout on region open. If a region open fails N times, put it aside.