Re: Cache Overfilling on SOLARIS

From: Jens-S. Voeckler <voeckler@dont-contact.us>
Date: Tue, 18 Jul 2000 10:18:37 +0200 (CEST)

On Mon, 17 Jul 2000, Murphy Terrance M wrote:

]--------------------------------------------------------
]diskHandleWrite: FD 6: disk write error: (28) No space left on device
]FATAL: Write failure -- check your disk space and cache.log
]Squid Cache (Version 2.3.STABLE3): Terminated abnormally.
]CPU Usage: 13.210 seconds = 3.160 user + 10.050 sys
]Maximum Resident Size: 0 KB
]--------------------------------------------------------
]I have read many messages concerning like problems in this forum but have
]not been able to cure my problem. The only thing that works is to newfs the
]cache directories and run squid -z and start over. It happens so often that

Ok, I think I know your problem - I assume that you did apply all
necessary NLANR patches to rule out Squid's fault in overfilling.

I believe it to be mentioned in the system-dependent weirdness section

        http://www.squid-cache.org/Doc/FAQ/FAQ-14.html

in the FAQ. You ran against the wall, because your UFS was time-optimized.

The reason is that the fragmentation fills the disk. Check, what the
first "fstyp -v <raw/dev> | grep ^nbfree | head -1" says. Probably it went
close to zero or even zero. If zero, there is no more free space on the
FS, regardless of what df/du tells you.

If nbfree went 0, it is not Squid's fault. You have the following options:

1) get Veritas FS. Please note that it will cost $$, but also
   might offer some other benefits. I haven't tried this option due to
   the notorious money shortage in education...
2) only use 50% of the disks. When using time optimization, fragmentation
   is guaranteed *not* to kill you, if you use only 50%. If you use any
   more of the disk with time-optimization, fragmentation might not kill
   you immediately, but believe me, it usually kills you Sunday night at
   03:45.
3) use space optimization. That is the easiest and less costly fix. If
   using modern drives, or E450 PCI-HW-RAID0, the difference is "almost
   undetectable" in many cases.

You haven't mentioned the Solaris version. There is supposed to be a bug
with 2.6 newfs, which makes wrong assumptions about disk layout. I believe
that was also mentioned in the FAQ.

Le deagh dhùrachd,
Dipl.-Ing. Jens-S. Vöckler (voeckler@rvs.uni-hannover.de)
Institute for Computer Networks and Distributed Systems
University of Hanover, Germany; +49 511 762 4726
Received on Tue Jul 18 2000 - 02:22:48 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:54:33 MST