squid 2.0.PATCH2 filled cache and stopped

The other night our squid cache (2.0.PATCH2) stopped at 1:19 in the morning, after filling one
of our cache directories to the hilt... the following error messages were found in cache.log:

1998/11/15 01:18:58| diskHandleWrite: FD 73: disk write error: (28) No space left on device
1998/11/15 01:18:58| storeSwapOutHandle: SwapOut failure (err code = -6).
1998/11/15 01:18:58| WARNING: Shrinking cache_dir #3 to 1903410 KB
1998/11/15 01:19:00| diskHandleWrite: FD 43: disk write error: (28) No space left on device
1998/11/15 01:19:00| storeDirWriteCleanLogs: Starting...
1998/11/15 01:19:00| WARNING: Closing open FD 66
1998/11/15 01:19:00| WARNING: Closing open FD 67
1998/11/15 01:19:00| storeDirWriteCleanLogs: /usr/local/squid/cache/4/swap.state.clean: write: (28) No
space left on device
1998/11/15 01:19:00| storeDirWriteCleanLogs: Current swap logfile not replaced.
1998/11/15 01:19:01| 65536 entries written so far.
1998/11/15 01:19:01| 131072 entries written so far.
1998/11/15 01:19:02| 196608 entries written so far.
1998/11/15 01:19:02| 262144 entries written so far.
1998/11/15 01:19:03| 327680 entries written so far.
1998/11/15 01:19:03| Finished. Wrote 379623 entries.
1998/11/15 01:19:03| Took 3 seconds (126541.0 entries/sec).
FATAL: Write failure -- check your disk space and cache.log

Every restart after this died after trying to write to the full directory :(

Perhaps squid could just disable the full directory in cases like these and
continue to use the remaining ones, rather than just stopping?

Regards,

Lachlan Cameron-Smith lok@box.net.au
BOX Internet Services Systems Administrator (08) 8211 8500
Received on Mon Nov 16 1998 - 17:26:18 MST