Hi Henrik,
Thx for the answer. I looked the network side, but it seems ok. I have ran a ping
between each proxy during 4 hours of hight HTTP load, with 0% of packet loss and
very good response time.
It 's probably more an overload issue. The load of every proxy is about 80-90% during
the "working hours" (monitored with mrtg). And since the beginning of the summer
holidays,
the HTTP load has significantly increased. I have 5 proxies chained on three parents
(each
son has 2 parents defined), and each of the 5 "son proxy" handles between 3000 and
4000
HTTP requests/min. These are dl360 compaq CPU 1Ghz with 1GB of RAM and scsi drives
(and the
same for the parent proxies) running on RH9 with squid 2.5 ; do you think the hardware
could
be the weak point?
Thx for your help.
Regards,
P-E
-----Message d'origine-----
De : Henrik Nordstrom [mailto:[EMAIL PROTECTED]
Envoy� : mercredi 4 ao�t 2004 12:21
� : SXB6300 Mailing
Cc : [EMAIL PROTECTED]
Objet : Re: [squid-users] Detected DEAD parent
On Wed, 4 Aug 2004, SXB6300 Mailing wrote:
> I have already increased the icp_query_timeout and maximum_icp_query_timeout
> parameters, but with no success. This problem has appeared recently with
> no changes on squid boxes. What could be the reason of it? The network?
Packet loss of ICP messages, and as you also see TCP Connection falied
warnings there is also more servere issues..
Either your parents is overloaded, or there is serious network
connectivity problems.
Regards
Henrik
