Hello All,
*Short version*: Looking to leverage hbase fuzzy row scans
on fields in a composite rowkey ( on hbase 0.94.6 ).
*Long version*: Our hbase store design is heavily dependent on a bucketed
composite rowkey that uses all the lovely features in hbase to spread the
read/write workload across regions. We want to do incremental processing
using Hive queries (over hbase of course) on new data entering the system.
It's of paramount importance to be able to reduce the scan size on each
iteration of the HQL jobs.
We are running hbase 94.6 with Hive 0.10. Upgrading to a later version of
hive is easy for us but upgrading Hbase will be much much harder. I have
gone through the current docs under
https://cwiki.apache.org/confluence/display/Hive/HBaseIntegration and
related jiras viz. HIVE-6290, HIVE-6411 et al . Both those are patch
available but possibly going into hive 0.13. Hive 0.13 is compatible with
hbase 96 only I believe. My current plan is to embark on a patching
expedition to try and retrofit these patches onto earlier versions of hive
that play nice with hbase 94.6 But i reckon: "here be dragons".
Ergo I am relying on the mighty hive user group to provide guidance in this
hour of need. Please lead me and I shall follow.
Cheers,
-Gautam.
====
*Footnote*: Our rowkey is a fixed length key(no delimiter) that looks like:
[bucket][App][MsgType][TimeSlice] ....
