[Lias] Thanks for help re Proxy

[Craig Ringer]

> Squid will make a filesystem call, and the OS will return an error code. 
> First year university students know how to check these codes, I doubt 
> that a package like squid wouldn't pay attention to them.

In my experience, squid dies silently when it runs out of disk space. As 
a result, the first warning you get is the users. Of course, I'm always 
running gkrellm etc and keeping an eye on free disk space using the snmp 
module for gkrellm, so I haven't had it happen in production use. Oh, 
execept once, where something went runaway and filled up /var with 
syslog messages.

Ideally, squid could email the admin saying "Arrggh, out of disk space 
in cach directory /var/cache/squid. Going to passthrough mode, will not 
add to cache until some space is freed."

> More than likely the disk is faulty, plain and simple. I would be 
> looking into adding a new drive, salvaging your data, and getting rid of 
> the old one (depending on how critical your data on that disk is).

Possible ... but he did mention a crash, probably a kernel panic by the 
sounds of things. Did it look a bit like this:

Unable to handle kernel NULL pointer dereference at virtual address 
00000002c
c026d0c9
*pde = 00000000
Oops: 0000
CPU:    0
EIP:    0010:[<c026d0c9>]    Not tainted
Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010206
eax: 00000028   ebx: dfccc3a8   ecx: 01800204   edx: 00000001
esi: dfccc3a8   edi: dfc265a0   ebp: 00000000   esp: defe5da8
ds: 0018   es: 0018   ss: 0018
Process syslogd (pid: 281, stackpage=defe5000)
Stack: c15d957c dfc265a0 c15d9560 00000000 df381018 c01da1dc dfccc3c0 
dfc26620
        dfccc3e0 dfccc380 dfc0c1e0 00000000 c026d750 c15d9560 dfc265a0 
00000001
        c15d957c c15d9560 00000001 00000000 00000008 c038e600 00000000 
c026d9c2
Call Trace:    [<c01da1dc>] [<c026d750>] [<c026d9c2>] [<c010985d>] 
[<c01099c6>]
   [<c010bbc8>] [<c02707e9>] [<c0271792>] [<c012f3e4>] [<c012f5c3>] 
[<c012f5de>]
   [<c01408fa>] [<c0140bf8>] [<c027188e>] [<c0271f44>] [<c010856f>]
Code: 8b 2c 90 8b 44 24 28 c1 e9 08 83 e1 0f d3 ed 83 e5 01 c7 44


because if so, it was probably a kernel panic that took it down. The 
reason for the file system corruption in this case is usually that the 
fs was in the middle of writing data to the disk when the machine crashed.

Journalling filesystems like reiserfs and ext3 are designed to help 
prevent this problem. If you have a vaguely recent distro with a 2.4 
kernel, you should be able to convert your existing filesystems from 
ext2 to ext3 relatively easily.... after you back up, just to make sure. 
There's plenty of info on the 'net about how to do this.

Craig Ringer
IT Manager
POST Newspapers