Thanks very much for the help. I'm currently using a RegionObserver (to
implement readless increments) on a subset of my tables, and placing the jar on
hdfs. I will try what Clay mentioned and see if it works.
Best,Albert
On Wednesday, November 23, 2016 4:42 AM, Clay Baenziger (BLOOMBERG/ 731
LEX) <[email protected]> wrote:
Hi Albert,
The way I've seen it working -- if you need to leave the table online -- is
specifying the coprocessor in the hbase-site.xml of each region server and
deploying the class to the table without a path (just the classname). Then,
ensure your hbase-site.xml has the correct version for the region server
running on that machine.
Esther and I briefly talked on this type of operation in
http://www.slideshare.net/EstherKundin/coprocessors-uses-abuses-solutions-presented-at-hbasecon-east-2016
or https://www.youtube.com/watch?v=dYeJd9AAVlw
-Clay
From: [email protected] At: 11/18/16 19:33:54To: [email protected],
[email protected]
Subject: Re: Rolling upgrade with coprocessors
Can you tell us the version of hbase you are using and the new version
which you plan to upgrade to ?
A bit more detail on your coprocessor would also help narrow the scope of
search.
Cheers
On Fri, Nov 18, 2016 at 4:28 PM, Albert Shau <[email protected]>
wrote:
> Hi all,
> I'm using coprocessors with my tables and am wondering how I would perform
> an HBase rolling upgrade, since it seems like there are no compatibility
> guarantees for the coprocessor APIs. I'm guessing I would have to disable
> the table, alter the table to use a coprocessor compatible with the next
> HBase version, perform the rolling upgrade, then enable the table? Is there
> a way to keep the table enabled for this?
> Thanks,Albert
