dnsserver dying

1234567890123456789012345678901234567890123456789012345678901234567890
I am running Squid 2.2_STABLE5 on Solaris 2.6 and am experiencing the
following problem. About every 15-20 minutes, I get this message:

        Nov 17 11:48:10 proxy squid[2681]: Too many queued dnsserver requests

Looking at the squid log files, I see the following:

        1999/11/17 11:48:10| WARNING: All dnsserver processes are busy.
        1999/11/17 11:48:10| WARNING: 15 pending requests queued
        1999/11/17 11:48:10| Consider increasing the number of dnsserver
                processes in your config file.
        FATAL: Too many queued dnsserver requests
        Squid Cache (Version 2.2.STABLE5): Terminated abnormally.
        CPU Usage: 195.310 seconds
        Maximum Resident Size: 0 KB
        Page faults with physical i/o: 1300
        1999/11/17 11:48:13| Starting Squid Cache version 2.2.STABLE5 for
                sparc-sun-solaris2.6...
        1999/11/17 11:48:13| Process ID 3842
        1999/11/17 11:48:13| With 1024 file descriptors available
        1999/11/17 11:48:13| Performing DNS Tests...
        1999/11/17 11:48:13| Successful DNS name lookup tests...
        1999/11/17 11:48:13| helperOpenServers: Starting 10 'dnsserver'
                processes

The machine has 2GB RAM and 4 CPU's, so I don't think it is a resource
problem. Is this a problem with STABLE5? I have STABLE4 running on a
lesser machine, and it handles about 3 times more traffic without any
problems. Both machines start up 10 dnsserver processes, but it appears
most of them don't get any use (ie. zero CPU time). I have also tried
increasing the number of dnsservers to 20, and that did nothing as well.
When it does die, it takes squid with it and promptly restarts the
entire
proxy.

Any suggestions/solutions would be welcome.

Thanks,
Jason
Received on Wed Nov 17 1999 - 11:10:50 MST