-Re: ANN: The 3rd hbase 0.94.2 release candidate is available for download
Gregory Chanan 2012-10-03, 00:22
I ran some YCSB tests on a cluster. I think the necessary factor is that
your cluster gets socket timeout exceptions -- there are a ton in the logs
of this cluster. I haven't tracked down what is causing them yet -- the
usual suspects are network flakiness or long GCs.
It doesn't occur in 0.92. I actually found that the cause was HBASE-5058
by binary searching through the 0.94 commits.
On Tue, Oct 2, 2012 at 5:09 PM, lars hofhansl <[EMAIL PROTECTED]> wrote:
> Thanks for testing Gregory,
> could (either here or in the jira) outline how you got into this situation?
> I loaded a billion rows into one of our test clusters and played around
> with that data and have not seen this issue.
> I'd also be very interested to know whether this existed in 0.92 (and even
> in 0.90).
> -- Lars
> From: Gregory Chanan <[EMAIL PROTECTED]>
> To: [EMAIL PROTECTED]; lars hofhansl <[EMAIL PROTECTED]>
> Sent: Tuesday, October 2, 2012 10:20 AM
> Subject: Re: ANN: The 3rd hbase 0.94.2 release candidate is available for
> I tried out the client (ok, not the actual client, but pretty close) on a
> benchmark app and it got stuck. I filed:
> Not sure if that necessitates a -1, because it appears to happen on earlier
> 0.94 versions as well.
> On Mon, Oct 1, 2012 at 10:50 PM, lars hofhansl <[EMAIL PROTECTED]>
> > Thanks Stack.
> > Indeed the api docs are missing (from the 0.94.1 tarball as well).
> > I need to check the jenkins build, which generates the tarballs.
> > -- Lars
> > ----- Original Message -----
> > From: Stack <[EMAIL PROTECTED]>
> > To: [EMAIL PROTECTED]; lars hofhansl <[EMAIL PROTECTED]>
> > Cc:
> > Sent: Monday, October 1, 2012 9:37 PM
> > Subject: Re: ANN: The 3rd hbase 0.94.2 release candidate is available for
> > download
> > On Mon, Oct 1, 2012 at 3:14 PM, lars hofhansl <[EMAIL PROTECTED]>
> > > Friendly reminder to please give this RC a spin.
> > >
> > I tried it locally. Updated the spreadsheet with what I did. Javadoc
> > is missing but I wouldn't sink the release for that.
> > +1.
> > St.Ack