Growing squid, fd_table error and slow storage rebuilding

Hello,

we're running squid on a Solaris 7 intel box, version 2.2.STABLE4 for
i386-pc-solaris2.6.

We have two problems with squid:

The squid is growing permanently, slowly during low traffic times, about 3 to 4
MB per hour during working hours. Used memory after one week running (from top):
SIZE RES STATE TIME CPU COMMAND
346M 333M run 513:36 1.92% squid

Does anybody know if this is a normal behaviour or is something going nuts here?

Second, we found following error message in cache.log:

comm_write: fd_table[143].rwstate != NULL

We discoverd this error message after a breakdown of squid (that means, squid
was still an accepted client requests, but didn't deliver anything to the
clients). It need not neccessarily be correlated...

Oh yes, and before I forget: Our squid's rebuilding it's storage at every time:

1999/11/10 11:05:27| Rebuilding storage in Cache Dir #0 (DIRTY)
1999/11/10 11:05:27| Rebuilding storage in Cache Dir #1 (DIRTY)
...
1999/11/10 11:50:06| Finished rebuilding storage disk.

Any hint is welcome.

Regards,

Joerg

// pallas GmbH ............ Joerg Oertel ...........
   Hermuelheimer Str. 10 System engineer
   D-50321 Bruehl, Germany Joerg.Oertel@pallas.com
                               phone +49-(0)2232-1896-0
   http://www.pallas.de fax +49-(0)2232-1896-29
........................................................
Received on Thu Nov 25 1999 - 08:35:26 MST