hg pull: out-of-memory errors? ("abort: group to be added is empty!")

hg pull: out-of-memory errors? ("abort: group to be added is empty!")

[Hollis Blanchard]

Hi James, I've been getting errors when pulling with hg 0.8.1. The
problem is apparently that the server process is being killed, and hg
has only recently started reporting that error. The error text is
"abort: group to be added is empty!", which the hg guys need to improve
(http://www.selenic.com/mercurial/bts/issue210).

Could you check the xenbits server logs to see if there are hg processes
being killed, possibly because of out-of-memory conditions? Thanks...

-- 
Hollis Blanchard
IBM Linux Technology Center

Re: hg pull: out-of-memory errors? ("abort: group to be added is empty!")

[Bryan O'Sullivan]

On Wed, 2006-04-19 at 14:04 -0500, Hollis Blanchard wrote:

> Could you check the xenbits server logs to see if there are hg processes
> being killed, possibly because of out-of-memory conditions? Thanks...

It may also be worth upgrading the server to the current tip version of
Mercurial, once that has seen some more testing.  Matt just merged some
changes written by Chris Mason that reduce disk and memory usage
substantially, which should reduce the squeeze that the xenbits server
gets put under.

I'll see what we can do to print a more useful error message when the
xenbits server gets killed, as it's quite a common problem.

	<b

-- 
Bryan O'Sullivan <bos@serpentine.com>

Re: hg pull: out-of-memory errors? ("abort: group to be added is empty!")

[James Bulpin]

There is nothing untoward in any log as far as I can see. The oom killer 
hasn't come out to play either.

Anyway, xenbits is now running 0.8.1, I'll upgrade as Bryan suggests 
once a new version is called.

We've had 0.8.1 testing on an internal server for a few days so I have 
confidence the upgrade should be safe. As ever, do let me know if any 
problems are seen.

Note that the temporary server that the 3.0-testing tree is redirected 
to is still running 0.7 but will disappear shortly once I move the 
hosting to our new colo centre.

Regards,
James

Bryan O'Sullivan wrote:
> On Wed, 2006-04-19 at 14:04 -0500, Hollis Blanchard wrote:
> 
> 
>>Could you check the xenbits server logs to see if there are hg processes
>>being killed, possibly because of out-of-memory conditions? Thanks...
> 
> 
> It may also be worth upgrading the server to the current tip version of
> Mercurial, once that has seen some more testing.  Matt just merged some
> changes written by Chris Mason that reduce disk and memory usage
> substantially, which should reduce the squeeze that the xenbits server
> gets put under.
> 
> I'll see what we can do to print a more useful error message when the
> xenbits server gets killed, as it's quite a common problem.
> 
> 	<b
>