Re: [squid-users] squid crashing consistently under high loads (2.4STABLE1)

Adrian:

Thanks for your reply. Currently, /proc/meminfo on one of the boxes shows:

MemTotal: 1028024 kB
MemFree: 2704 kB
MemShared: 0 kB
Buffers: 22572 kB
Cached: 58444 kB
Active: 61840 kB
Inact_dirty: 12392 kB
Inact_clean: 6784 kB
Inact_target: 792 kB
HighTotal: 131008 kB
HighFree: 1048 kB
LowTotal: 897016 kB
LowFree: 1656 kB
SwapTotal: 1542200 kB
SwapFree: 1509828 kB

--
Adam Woodbridge
----- Original Message -----
From: "Adrian Chadd" <adrian@squid-cache.org>
To: "Adam Woodbidge" <adam@woodbridge.ca>
Cc: <squid-users@squid-cache.org>
Sent: Tuesday, May 08, 2001 2:46 AM
Subject: Re: [squid-users] squid crashing consistently under high loads
(2.4STABLE1)
> On Mon, May 07, 2001, Adam Woodbidge wrote:
>
> [snip]
>
> > Mysteriously, squid doesn't report anything seriously wrong in cache.log
> > just before the box crashes.  However, after just recently enabling
kernel
> > console logging to the serial port, I saw the following error message
> > repeatedly displayed while the problem was happening on one of the
boxes:
> >
> > __alloc_pages: 1-order allocation failed.
> >
> > If I understand this message correctly, the kernel as failed to allocate
> > memory.  But how can this be when, even during peak usage, the squid
process
> > uses only about 300MB?  Each box has over 1GB of RAM in it, plus another
> > 1.5GB in swap space (which I've never seen used)!
>
> Your kernel also requires memory for things like socket buffers,
> packet buffers, tcp control blocks..
>
> It looks like you are running out of RAM in the kernel.
>
> > Any information or suggestions that could be provided towards resolving
this
> > problem would be very much appreciated.  I'd be happy to forward more
> > information if required.
>
> Whats the output of cat /proc/meminfo ?
>
> >
> > Regards,
> >
> > Adam Woodbridge
> >
>