You need to add lots of swap space. Not because it will ever really be needed,
but because Solaris (and most other UNIX:es) temporarily allocates lots of
swap space when a huge application starts subprocesses.
A rule of thumb is to have at least 3 times the size of your Squid process in
swap space.
Some UNIX:es solve this by allowing one to "overcommit" memory usage by
ignoring memory reservations and only counting what is really used. Not
enabled by default as it changes how "out of memory" is reported to the
applications if you really runs out of memory.
Regards
Henrik
Michael K Bender wrote:
> My squid cache, which is running on a Sun Ultra 10 with 512M RAM,
> running squid 2.4STABLE1 crashes almost every other day. I rotate the
> logfiles every weeknight. After rotating the logfiles, I get these
> messages:
>
> 2002/06/04 00:30:27| ipcCreate: fork: (12) Not enough space
> 2002/06/04 00:30:27| WARNING: Cannot run
> '/local/squid-2.4.STABLE1/libexec/squid/dnsserver' process.
>
> This is repeated for every dnsserver it tries to start, and the same for
> my auth processes. After doing this, I get the following:
>
> FATAL: Too many queued authenticator requests
> Squid Cache (Version 2.4.STABLE1): Terminated abnormally.
>
> After this it restarts and keeps going. I'm not sure if this is
> actually something I should be worried about. If it is, does anybody
> know why it is doing this and how to fix it.
>
> Mike
Received on Wed Jun 05 2002 - 08:01:39 MDT
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:08:27 MST