<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
14 out of 16G should be fine provided the machine isn't doing anything
else and the number of connections is under a few thousand. We run a
similar configuration.<br>
<br>
Also, attaching to the process with gdb and using the "where" command
will most likely be more useful than strace, especially if the program
is caught in an infinite loop and isn't making any system calls (in
which case strace won't output anything at all.)<br>
<br>
-Steve<br>
<br>
<br>
Jeremy wrote:
<blockquote cite="midC17C3920.66438%25jeremy@smokehabanos.com"
type="cite">
<pre wrap="">Next time it happens I can get an strace out of the process but that could
be a couple weeks until it occurs again.
Is there any issue with the amount of memory I'm dedicating? The machine
has a total of 16gigs and I'm giving 14 of that to memcached.
In our older machines, running an older version of memcached, we run
multiple instances. Is there any advantage to that? I was hoping to get
away from having to run multiple instances.
Thanks
-jeremy
On 11/12/06 1:24 AM, "Steven Grimm" <a class="moz-txt-link-rfc2396E" href="mailto:sgrimm@facebook.com"><sgrimm@facebook.com></a> wrote:
</pre>
<blockquote type="cite">
<pre wrap="">Can you attach a debugger to it and get a stack trace? It would help to
know where in the code it's hanging/looping.
-Steve
Jeremy wrote:
</pre>
<blockquote type="cite">
<pre wrap="">We've been having an issue with memcached 1.2.0 on Fedora Core 5 on am 64bit
AMD with 16gigs of memory.
Memcached is running with the following parameters:
memcached -d -p 11211 -u nobody -c 1024 -m 14336
About every two weeks we'll run into a situation where memcached will lock
up with 100% cpu utilization and become unresponsive. Only work around is
to restart the daemon at which point we loose our cache.
Is there any known issues around this? The interesting thing from our side
is that this has happened on the same exact machine 4 times now, while we
have another machine in the same cluster that has been running with no
problems at all. Exact same configuration. Nothing from the OS indicates a
hardware problem.
Please let me know what other information is important in helping to debug
this problem.
Thank You
-jeremy
</pre>
</blockquote>
</blockquote>
<pre wrap=""><!---->
</pre>
</blockquote>
<br>
</body>
</html>