Lily rowlog-mq Table Region Splits after initial start-up

[jim soya]

Our initial configuration was a 1 region rowlog-mq Hbase Table.

We are indexing about 10 million records per day.

We have 18 region servers in our Hadoop Cluster.

 

We had an issue with Lily not being able to keep up and ended up with 81 Million records in our rowlog-mq table.

We decided to split the rolwog-mq table into 16 regions to improve scan performance.

This appears to have worked for 20 Million of the records.

There are now 61 million records that are never pulled off of the rowlog-mq table.

New records are processed as expected.

 

1) Could this be due to the splitting of the rowlog-mq table?

2) Can we shutdown hbase and merge the rowlog-mq table back together?
 "$ bin/hbase org.apache.hadoop.hbase.util.Merge <tablename> <region1> <region2>"

Thanks,

Jim