Problems with parents under heavy load

Hello evryone,

I have squid 1.1.20 on Solaris 2.5.1 with the following configuration

cache_host leipzig.www-cache.dfn.de sibling 8080 3130
cache_host nuernberg.www-cache.dfn.de sibling 8080 3130
cache_host muenchen.www-cache.dfn.de sibling 8080 3130
cache_host stuttgart.www-cache.dfn.de sibling 8080 3130
cache_host wwwcache.leo.org sibling 8009 3130 proxy-only
cache_host proxy2.lrz-muenchen.de sibling 8080 3130 proxy-only weight=2

cache_host_domain nuernberg.www-cache.dfn.de .com
cache_host_domain stuttgart.www-cache.dfn.de .com
cache_host_domain leipzig.www-cache.dfn.de !.com
cache_host_domain muenchen.www-cache.dfn.de !.com

The problem is that muenchen.www-cache.dfn.de is reachable with ICP under
heavy load, but TCP requests are rejected and my proxy waits for objects
from this parent (probably that's the well known filehandle-problem under
Solaris and the parent is not tuned to use more than 1024 filehandles).
Although the cachemgr detects that the parent is down, the HTTP-requests
are forwarded to this reighbor. After I have configured this parent as
sibling the problem does not occur no more. My questions:

1. Is it possible to configure squid with a TCP-Timeout for parents, so
that after this timeout the query is forwarded to another parent (or
direct)?

2. or is this a Bug in one of the two squids (the parent is running with
squid 1.1.16 on Solaris 2.5.1)

------------------------------------------------------------------
      .---. .----------- Leibniz-Rechenzentrum Muenchen
     / \ __ / ------ Abt. Kommunikationsnetze
    / / \(..)/ -----
   ////// ' \/ ` --- Karim Tripodoro
  //// / // : : --- Tel.: +49 89 289-28839
 // / / /` '-- eMail: Karim.Tripodoro@lrz.de
// //..\\
------=====UU====UU======-----------------------------------------
           '//||\\`
             ''``
Received on Wed Jul 15 1998 - 00:32:42 MDT