Large clone chewing CPU

[Eric M. Hopper]

On Wed, 2005-08-31 at 00:38 -0700, Bryan O'Sullivan wrote:
> On Tue, 2005-08-30 at 18:15 -0700, Eric Hopper wrote:
> > 
> > During those periods, the hg process on the source box seems to largely
> > be doing a lot of mmaps and munmaps interspersed with writes of
> > seemingly repetititve garbage to stdout.
> 
> That's probably a combination of zlib and the wire protocol.

I'm going to be doing some investigation.  This has been going for 16
hours or so now.  There seem to be long periods of CPU chewing (lots of
that spent in the kernel because of the mmaps) with short periods of
time where actual data is transmitted.  I may be wrong about this, but
my guess is that the cycles correspond to changesets.

Oh, the mmaps appear to be mapping MAP_ANONYMOUS pages...

mmap2(NULL, 11440128, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS,
-1, 0)

Probably doesn't help much to know that though.  Which there was a way
to break into a Python process and figure out what the interpreter's
doing.

> > What's going on?
> 
> Hard to say why there's misbehaviour.  Is the repo in question something
> that outsiders could take a look at?

Sadly, no.  None of my repositories are that large.  And this one is all
proprietary code and jars and gifs and stuff.  One of these days I'll
get a job doing Open Source.  :-)

Have fun (if at all possible),
-- 
The best we can hope for concerning the people at large is that they
be properly armed.  -- Alexander Hamilton
-- Eric Hopper (hopper at omnifarious.org  http://www.omnifarious.org/~hopper) --
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 185 bytes
Desc: This is a digitally signed message part
Url : http://www.selenic.com/pipermail/mercurial/attachments/20050831/923002a1/attachment.pgp