[squid-users] Disk full on cache-disk

Hi,

We are running a squid 2.3.Stable4 on a Solaris 8 box (Sun Ultra 1) with 8 2Gb disks as cache disks. Logfiles and cache are on separate disks and only squid has write access to cache disks. We have had two incidents during the last 6 months when one or more cache disks has reported full i.e. file system full (physical disk ~ 1.9 Gb , cache dir size limit is 1.6 Gb, cache dir type ufs). Last time this occurred propably when a user downloaded CD-images (3x600Mb). Average disk usage is normally about 75 % (file system) but as an aftermath I noticed a cahce disk with 90 % usage (df -kl).

Normally there is not a problem with disk space , replacement and swapping out old entries seems to work , but on these occations at least one disk became full and squid died.

2002/03/07 09:23:47| diskHandleWrite: FD 230: disk write error: (28) No space left on device
2002/03/07 09:23:47| storeUfsWriteDone: got failure (-6)
2002/03/07 09:23:47| WARNING: Shrinking cache_dir #7 to 1408141 KB
2002/03/07 09:23:51| diskHandleWrite: FD 18: disk write error: (28) No space left on device
2002/03/07 09:23:51| storeDirWriteCleanLogs: Starting...
2002/03/07 09:23:51| WARNING: Closing open FD 6
2002/03/07 09:23:51| WARNING: Closing open FD 8
2002/03/07 09:23:51| storeDirWriteCleanLogs: /squid/d67/swap.state.clean: write: (28) No space left on device
2002/03/07 09:23:51| storeDirWriteCleanLogs: Current swap logfile not replaced.
FATAL: Received Segment Violation...dying.
2002/03/07 09:23:51| storeDirWriteCleanLogs: Starting...
2002/03/07 09:23:51| WARNING: Closing open FD 6
2002/03/07 09:23:51| WARNING: Closing open FD 8
2002/03/07 09:23:51| storeDirWriteCleanLogs: /squid/d67/swap.state.clean: write: (28) No space left on device
2002/03/07 09:23:51| storeDirWriteCleanLogs: Current swap logfile not replaced.

later on in startup:

2002/03/07 09:24:05| diskHandleWrite: FD 20: disk write error: (28) No space left on device
FATAL: Write failure -- check your disk space and cache.log
Squid Cache (Version 2.3.STABLE4): Terminated abnormally.
CPU Usage: 1.990 seconds = 1.220 user + 0.770 sys
Maximum Resident Size: 0 KB
Page faults with physical i/o: 0

and from messages file:

Mar 7 09:23:47 proxy1 ufs: [ID 845546 kern.notice] NOTICE: alloc: /squid/d67: file system full
Mar 7 09:23:51 proxy1 last message repeated 1 time
Mar 7 09:23:59 proxy1 ufs: [ID 213553 kern.notice] NOTICE: realloccg /squid/d67: file system full
Mar 7 09:23:59 proxy1 squid[19508]: [ID 702911 local4.alert] Write failure -- check your disk space and cache.log
Mar 7 09:24:05 proxy1 ufs: [ID 213553 kern.notice] NOTICE: realloccg /squid/d67: file system full
Mar 7 09:24:05 proxy1 squid[19510]: [ID 702911 local4.alert] Write failure -- check your disk space and cache.log
Mar 7 09:24:11 proxy1 ufs: [ID 213553 kern.notice] NOTICE: realloccg /squid/d67: file system full
Mar 7 09:24:11 proxy1 squid[19512]: [ID 702911 local4.alert] Write failure -- check your disk space and cache.log
Mar 7 09:24:17 proxy1 ufs: [ID 213553 kern.notice] NOTICE: realloccg /squid/d67: file system full
Mar 7 09:24:17 proxy1 squid[19514]: [ID 702911 local4.alert] Write failure -- check your disk space and cache.log
Mar 7 09:24:22 proxy1 ufs: [ID 213553 kern.notice] NOTICE: realloccg /squid/d67: file system full
Mar 7 09:24:22 proxy1 squid[19516]: [ID 702911 local4.alert] Write failure -- check your disk space and cache.log
Mar 7 09:24:22 proxy1 squid[366]: [ID 567784 local4.alert] Exiting due to repeated, frequent failures

From my point of view with the current config and limits the cache disks should never become full due to squid operations. But somehow they did and perhaps someone has an idea of why and how to avoid it ? I should mention that I will upgrade to 2.4.Stable4 asap.

Thanks,

Anders
Received on Fri Mar 15 2002 - 03:28:10 MST