LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: LVS/Apache cluster freezes from time to time

To: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: LVS/Apache cluster freezes from time to time
From: Jan Abraham <jan_abraham@xxxxxxx>
Date: Fri, 31 Oct 2003 16:16:19 +0100
Hi Joe, hi Jacob,

thanks for your ideas!

> The original suggestion still sounds best to me. I'd log the swap space used
> and see if it gets close to the limit when it freezes. Unfortunately 
> you don't have access to the amount of memory in use (AFAIK) as it's
> all allocated even if not being actively used. Disk I/O and load average
> are also available in `top`. Presumably they're being read from somewhere
> in /proc.

'top' and the 'vmstat' Nicolas suggested shows no or less swap activity
during all the time. See this 'vmstat 1' dump. There are just 52 kB
swapped out after/during the freeze.

           procs -----------memory---------- ---swap--
           r   b   swpd   free   buff  cache   si   so
12:57:15:  0   7   8004  20732 179984 512408    0    0
12:57:32: 40 113   8004  20828 179996 512412    0    0
12:57:34: 14   3   8132   9780 178044 502352    0    4
12:57:35:  4   0   8132  10740 178004 502132    0   52

The machines have 1GB. Apache uses around 300MB, 200MB is used for
buffers, and the remaining part for file caching. Swap (512MB) is nearly
unused.

I just tried to 'swapoff -a' - nothing happend, see the vmstat:

           procs -----------memory---------- ---swap--
           r   b   swpd   free   buff  cache   si   so
16:01:44:  1   0      0 117884 193316 512208    0    0
16:01:45:  0   0      0 117660 193320 512268    0    0
16:01:46:  0   0      0 117400 193344 512332    0    0
16:02:05: 47  45      0 117252 193376 512416    0    0
16:02:07: 13   1      0 108580 193528 513100    0    0
16:02:08:  4   0      0 108064 193588 513320    0    0

Same thing, except it now misses the significant decreasing cache use
after the freeze... ummmh... I guess Linux has also cached some of the
swap space and VM unit decided to swap it out during the freezes (but
this seems to be a side effect of the freeze, not the cause).


Regards,
Jan


<Prev in Thread] Current Thread [Next in Thread>