Memcached (1.1.12) Hanging in itmes.c:153

[Dean Michael C. Berris]

> -----Original Message-----
> From: Steven Grimm [mailto:sgrimm at facebook.com]
> Sent: Thursday, April 26, 2007 1:07 AM
> To: Dean Michael C. Berris
> Cc: memcached at lists.danga.com
> Subject: Re: Memcached (1.1.12) Hanging in itmes.c:153
> 
> It doesn't make any sense that it'd actually hang there. There's a
loop
> in process_command, but given the arguments I see are getting passed
> into that function, I don't see any way that would loop either.
> 

Yes, it doesn't make sense. I looked at the code myself, and it's just
an assignment to a pointer, inside an if(...).

> Is this something you can catch in the act, or is your stack trace
from
> a core file? If you're catching it live, I'd be very interested in
> seeing what it looks like if you single-step through the code,
stepping
> until it gets back to wherever you interrupted it (assuming there's an
> infinite loop going on here). Looking at the 1.1.12 code I don't see
how
> it could get stuck but obviously it can since you're seeing it!
> 

Believe it or not, I'm catching this in the act. I tried to do a "step",
but the step just hangs there -- nothing happening, nothing doing.

> This is not something we have ever seen on any version of memcached to
> my knowledge.
> 

Uh oh... Not so good news...

We're running it on a dual Xeon, with 4GB ram, the instance allocated to
use 640 MB. Maybe an issue with pointer sizes on 64-bit machines?

--
Dean Michael Berris
<dmberris at friendster.com>
Mobile: +639287291459
YMID: mikhailberis