hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Paul Cavallaro <paulcavall...@gmail.com>
Subject HBASE-6295
Date Sat, 11 Aug 2012 22:35:16 GMT
Hey All,

I've been working on taking a stab at HBASE-6295 (
https://issues.apache.org/jira/browse/HBASE-6295), which is about changing
the batching and write buffering of the client. The main gist is to change
the write buffer to not be per table but to be per region, and to remove
some unnecessary object creation to streamline the process as well. When I
started working on it though I noticed that were were a lot of comments
about how the Process class should be moved into HTable, and also that the
HConnectionImplementation's processBatchCallback is now deprecated in favor
of HTable's (which actually delegates back to the HConnectionImplementation
for the time being).

I was wondering then, how this might affect my implementation, and if it is
acceptable to continue to add/change functionality in both the Process
class and HConnectionImplementation as well. I apologize if this email is
too specific for this mailing list, and I would have posted it to JIRA, but
it's been down for the past two days, and I guess I'm just getting a little



  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message