hbase-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From iain wright <iainw...@gmail.com>
Subject Re: Are coprocessor really bad?
Date Sat, 15 Mar 2014 22:37:36 GMT
Pretty sure it depends on your implementation.

Its powerful but you have the ability to shoot yourself in the foot.

One example is, we tried watching post put on a table to increment values
in another table. This ended up causing deadlocks when regionservers tried
to pull an RPC to write to each other. We have followed as a general rule
to operate only on the region or regionserver the request originated on for
mutations. We moved the increments up the stack and just pass them in over
flume directly to the analytics table.
On Mar 15, 2014 2:41 PM, "Jean-Marc Spaggiari" <jean-marc@spaggiari.org>

> This caught my attention on the LinkedIn HBase group:
> "Coprocessors are not well thought out or implemented in HBase. "
> I'm not a coprocessors user. Never used them nor looked at them. So I can
> not comment.
> Are they really that bad? I never heard anything against coprocessors so
> far, so I'm wondering what's the status of this statement?
> I don't want to start a troll, just want to understand where this stand.
> JM

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