timeouts

From: Javier Puche. CSIC RedIRIS <[email protected]>
Date: Mon, 19 Jan 1998 20:11:37 +0100

Hi all,

 Running Squid 1.1.18 on a Sun Ultra-2 running Solaris 2.6 I am having
the following problem:

 Some requests return a timeout error, the thing is that Squid returns
the timeout error in a very little time (about 10 secconds or less).
The directive for timeout is set to about two minutes so the only
reason I suspect is that the Operating System TCP stack has some
parameter that needs tuning for being more permisive with timeouts, I
am trying to find it but if any of you has had any experience with that
I thank the help. Any help from people who had also suffered timeouts
problems will be great, the problem might be other that what I was
pointing out.

 Our links with USA have very high packet loss, but going directly
instead of through the proxy, the document is normally succesfully
retrieved (though pretty slowly)

 Browsing today's log I see 14000 TIMEOUT_DIRECT (out of 1 million
TCP+UDP requests, only 169236 external fetches) and Calamaris
statistics show that almost 7% requests are ERR_CONNECT_FAIL; there
are few ERR_READ_TIMEOUT ... by the way, where can I find info on the
meaning of each ERR code ??

 I seem to remember timeout discussions in this list but I am unable to
locate them in recent months archives :-( (neither with the nice search
tool at cineca)

Thanks for any help !

Regards,

Javier Puche.
Received on Mon Jan 19 1998 - 11:18:34 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:38:27 MST