Not sure how to solve it manually but there seems to be a back end job to auto rebuild index.

Refer to Failure policy on http://phoenix.apache.org/secondary_indexing.html

-Dhaval


On Fri, Feb 27, 2015 at 9:53 PM, Jude K <j2k850@gmail.com> wrote:
Hi,

Been stuck on this issue for few hours. Hoping someone can shed some light.

OS: Centos-6
Phoenix Client: phoenix-4.0.0.2.1.5.0-695-client.jar
Phoenix Core: phoenix-core-4.0.0.2.1.5.0-695.jar
Hbase Version: Version 0.98.0.2.1.5.0-695-hadoop2, 6 RS
Hbase RS Java Heap : 6 Gb

So,

1) Created a 1 column family, 6 column Phoenix table
2) Have an app that is continually streaming  data into the new Phoenix table
3) Created a Phoenix index on two of the columns of the new Phoenix. 
4) Compare newly created Phoenix index count to row count in Hbase. They agree.
5) Wait a few minutes, do another comparison between Phoenix index count and Hbase row count. Row count properly increments, but Phoenix index count shows same value.
6) Wait some more and get exact same outcome as in #5.
7) Check a RS log file, and see that the Phoenix index is "DISABLED" because it can not write to a particular region.

OK.. what would cause the Phoenix index to become "DISABLED", especially since there were no issues during index creation? Is there a configuration variable that needs to be modified? Is there a suitable workaround beside developing a bash script to alter the index and rebuild.


Thanks