[squid-users] Help Required : No buffer space available

Hello,

I'm running Squid-2.4.STABLE6 (in accelerator mode) on FreeBSD 4.4-RELEASE
(dual P3 450Mhz)

After 5 minutes of running, I'm getting this error in my cache.log:

comm_open: socket failure: (55) No buffer space available.

And then the problem just seems to get worse and worse. The system doesn't
appear to have a huge load (no more than 4% CPU) with about 100 users
navigating the site at any one time (perhaps 5-10 requests made to the
server every second)

Here's an except from the cache.log

2002/04/15 21:38:28| Starting Squid Cache version 2.4.STABLE6 for
i386-unknown-freebsd4.4...
2002/04/15 21:38:28| Process ID 56543
2002/04/15 21:38:28| With 1064 file descriptors available
2002/04/15 21:38:28| Performing DNS Tests...
2002/04/15 21:38:28| Successful DNS name lookup tests...
2002/04/15 21:38:28| DNS Socket created on FD 4
2002/04/15 21:38:28| Adding nameserver 195.60.0.1 from /etc/resolv.conf
2002/04/15 21:38:28| Unlinkd pipe opened on FD 9
2002/04/15 21:38:28| Swap maxSize 102400 KB, estimated 7876 objects
2002/04/15 21:38:28| Target number of buckets: 393
2002/04/15 21:38:28| Using 8192 Store buckets
2002/04/15 21:38:28| Max Mem size: 16384 KB
2002/04/15 21:38:28| Max Swap size: 102400 KB
2002/04/15 21:38:28| Rebuilding storage in /apache/squid/cache (DIRTY)
2002/04/15 21:38:28| Using Least Load store dir selection
2002/04/15 21:38:28| Set Current Directory to /apache/squid/cache
2002/04/15 21:38:28| Loaded Icons.
2002/04/15 21:38:28| Accepting HTTP connections at 0.0.0.0, port 80, FD 11.
2002/04/15 21:38:28| WCCP Disabled.
2002/04/15 21:38:28| Ready to serve requests.
2002/04/15 21:38:29| Done reading /apache/squid/cache swaplog (835 entries)
2002/04/15 21:38:29| Finished rebuilding storage from disk.
2002/04/15 21:38:29| 607 Entries scanned
2002/04/15 21:38:29| 0 Invalid entries.
2002/04/15 21:38:29| 0 With invalid flags.
2002/04/15 21:38:29| 379 Objects loaded.
2002/04/15 21:38:29| 0 Objects expired.
2002/04/15 21:38:29| 228 Objects cancelled.
2002/04/15 21:38:29| 0 Duplicate URLs purged.
2002/04/15 21:38:29| 0 Swapfile clashes avoided.
2002/04/15 21:38:29| Took 0.9 seconds ( 409.3 objects/sec).
2002/04/15 21:38:29| Beginning Validation Procedure
2002/04/15 21:38:29| Completed Validation Procedure
2002/04/15 21:38:29| Validated 379 Entries
2002/04/15 21:38:29| store_swap_size = 2010k
2002/04/15 21:38:29| storeLateRelease: released 0 objects
2002/04/15 21:41:06| parseHttpRequest: Unsupported method 'User-Agent:'
2002/04/15 21:41:06| clientReadRequest: FD 72 Invalid Request
2002/04/15 21:41:06| clientSendMoreData: Deferring
error:unsupported-request-method
2002/04/15 21:49:09| Restarting Squid Cache (version 2.4.STABLE6)...
2002/04/15 21:49:09| FD 11 Closing HTTP connection
2002/04/15 21:49:09| DNS Socket created on FD 4
2002/04/15 21:49:09| Adding nameserver 195.60.0.1 from /etc/resolv.conf
2002/04/15 21:49:09| Accepting HTTP connections at 0.0.0.0, port 80, FD 8.
2002/04/15 21:49:09| WCCP Disabled.
2002/04/15 21:49:09| Loaded Icons.
2002/04/15 21:49:09| Ready to serve requests.
2002/04/15 21:54:25| comm_open: socket failure: (55) No buffer space
available
2002/04/15 21:54:25| comm_open: socket failure: (55) No buffer space
available
...
... (another 100 lines of the same error)
...
2002/04/15 22:11:57| comm_open: socket failure: (55) No buffer space
available
2002/04/15 22:11:59| comm_open: socket failure: (55) No buffer space
available
2002/04/15 22:12:12| Preparing for shutdown after 26138 requests
2002/04/15 22:12:12| Waiting 30 seconds for active connections to finish
2002/04/15 22:12:12| FD 8 Closing HTTP connection
2002/04/15 22:12:44| Shutting down...
2002/04/15 22:12:44| Closing unlinkd pipe on FD 9
2002/04/15 22:12:44| storeDirWriteCleanLogs: Starting...
2002/04/15 22:12:44| Finished. Wrote 679 entries.
2002/04/15 22:12:44| Took 0.1 seconds (7277.6 entries/sec).
CPU Usage: 75.005 seconds = 43.029 user + 31.976 sys
Maximum Resident Size: 8532 KB
Page faults with physical i/o: 0
2002/04/15 22:12:44| Squid Cache (Version 2.4.STABLE6): Exiting normally.

After I'd killed Squid, I did a netstat -mb and got:

78/912/4096 mbufs in use (current/peak/max):
        78 mbufs allocated to data
70/894/1024 mbuf clusters in use (current/peak/max)
2016 Kbytes allocated to network (65% of mb_map in use)
0 requests for memory denied
0 requests for memory delayed
0 calls to protocol drain routines

Is there any other info you need ?
Received on Tue Apr 16 2002 - 04:54:29 MDT