How to prevent a long swap.state rebuild

Hi,

Some time ago (1 month, actually) I sent the following message. I don't
think it made it to the list (I was posting via my mail/news gateway) so
I've re-submitted it.

The reason for the renewed interest is that I've had exactly the same
problem again. I've shut down my cache box to check the BIOS and, when I
restarted, it began re-building swap.state and is still going...

I use "shutdown -h now" to bring down the FreeBSD box. Should I shutdown
squid first?

Any ideas how I can prevent this from happening again in the future?

Thanks,

R.
======original message=======
I've got squid 2.1.PATCH2 running on a FreeBSD 3.1-RELEASE box. The
machine is a P133, (was 96M) 128M RAM, with a 1G system disk (256M swap)
and 8G cache disk, although I've limited the cache size to 4GB. I've
got soft-updates on both disks.

It seemed to be working OK when I first installed but I've recently
shutdown the box to replace the RAM - I've upped it to 128M.

Everything came back up OK but it took forever to rebuild
swap.state. I suspect that this is because the squid process is 145M in
size and I've only got 128M of RAM so there was continual swapping.

Here's cache.log:

1999/04/20 14:06:57| Ready to serve requests.
1999/04/20 14:06:58| Configuring Parent cache.www.uk.psi.net/8080/3130
1999/04/20 14:07:02| 65536 Entries read so far.
1999/04/20 14:07:07| 131072 Entries read so far.
1999/04/20 14:07:13| 196608 Entries read so far.
1999/04/20 14:07:20| 262144 Entries read so far.
1999/04/20 14:07:25| 327680 Entries read so far.
1999/04/20 14:07:28| 393216 Entries read so far.
1999/04/20 14:07:30| 458752 Entries read so far.
1999/04/20 14:07:32| 524288 Entries read so far.
1999/04/20 14:07:46| 589824 Entries read so far.
1999/04/20 14:08:06| 655360 Entries read so far.
1999/04/20 14:08:27| 720896 Entries read so far.
1999/04/20 14:08:50| 786432 Entries read so far.
1999/04/20 14:09:14| 851968 Entries read so far.
1999/04/20 14:09:40| 917504 Entries read so far.
1999/04/20 14:10:07| 983040 Entries read so far.
1999/04/20 14:10:35| 1048576 Entries read so far.
1999/04/20 14:11:05| 1114112 Entries read so far.
1999/04/20 14:11:37| 1179648 Entries read so far.
1999/04/20 14:12:09| 1245184 Entries read so far.
1999/04/20 14:12:43| 1310720 Entries read so far.
1999/04/20 14:13:19| 1376256 Entries read so far.
1999/04/20 14:13:58| 1441792 Entries read so far.
1999/04/20 14:14:38| 1507328 Entries read so far.
1999/04/20 14:15:27| 1572864 Entries read so far.
1999/04/20 14:19:15| 1638400 Entries read so far.
1999/04/20 16:56:22| 1703936 Entries read so far.
1999/04/20 22:39:15| 1769472 Entries read so far.
1999/04/21 06:47:12| 1835008 Entries read so far.
1999/04/21 19:53:30| 1900544 Entries read so far.
1999/04/22 00:05:04| Done reading Cache Dir #0 swaplog (3313667 entries)
1999/04/22 00:05:13| Finished rebuilding storage disk.
1999/04/22 00:05:13| 1923023 Entries read from previous logfile.
1999/04/22 00:05:13| 0 Entries scanned from swap files.
1999/04/22 00:05:13| 0 Invalid entries.
1999/04/22 00:05:13| 0 With invalid flags.
1999/04/22 00:05:13| 267712 Objects loaded.
1999/04/22 00:05:13| 0 Objects expired.
1999/04/22 00:05:13| 1343451 Objects cancelled.
1999/04/22 00:05:13| 2130 Duplicate URLs purged.
1999/04/22 00:05:13| 309730 Swapfile clashes avoided.
1999/04/22 00:05:13| Took 122296 seconds ( 2.2 objects/sec).
1999/04/22 00:05:13| Beginning Validation Procedure
1999/04/22 00:15:29| 65536 Entries Validated so far.
1999/04/22 00:15:31| 131072 Entries Validated so far.
1999/04/22 00:15:32| 196608 Entries Validated so far.
1999/04/22 00:15:33| 262144 Entries Validated so far.
1999/04/22 00:15:33| Completed Validation Procedure
1999/04/22 00:15:33| Validated 268529 Entries
1999/04/22 00:15:33| store_swap_size = 2322784k
1999/04/22 00:15:48| storeLateRelease: released 4825 objects

122296 seconds is just under 34 hours!!! During this time, client
performance was awful.

Now the rebuld has finished, the cache seems to be performing OK again.

Is this normal behaviour? Should this file grow so large? Will the
squid process return to normal size once this file is built? Are there
any configuration parameters I can tweak to reduce the size of this
file?

Thanks for any suggestions/advice/flames <g>

R.
 --
Robin Bowes, System Development Manager, Equal Opportunities Commission,
Room 405A, Overseas House, Quay St., Manchester, M3 3HN, UK.
Tel: +44 (0) 161 838 8321 Fax: +44 (0) 161 835 1657
Received on Fri May 21 1999 - 03:35:47 MDT