[squid-users] 504 Timeout of large dynamically generated files - delay pool problem?

From: Adam <adam-s@dont-contact.us>
Date: Fri, 13 Jun 2003 18:23:02 -0700

Hello,

We are having a problem with a ColdFusion generated report http page timing
out. The server is offsite and internal users connect to it via Squid which
throttles bandwidth with a delay pool. The ColdFusion generated pages do
take a while to generate (usually 5 minutes) but when I unproxy my PC, the
page comes up ok (after 5 or so minutes of generating the report). The
reported page "size" (per the IE download manager which pops up to move the
file from it's temp location to the correct location) states the file is
1.2MB but the actual size when downloaded seems to be around 25K. However
the file is a spreadsheet - that is it automagically calls Excel and opens
into a spreadsheet.

This worked on the Squid proxy prior to our implementing delay_pools.
However now that we have enabled delay pools, it breaks. Making the delay
pool bucket larger or the bandwidth given larger isn't our ideal option so I
was hoping some of you have a better idea. Also the application is only
http, no other protocol.

Here is the error as returned by the access log:
Fri Jun 13 12:20:15 2003 300028 192.168.5.4 TCP_MISS/504 1396 POST
http://external.web.server/A_44000.cfm/QueryFacility.xlc? -

Here is what appears on the browser:
DIRECT/209.157.104.251 text/html

While trying to retrieve the URL: [no URL]
              The following error was encountered:
Read Timeout
               The system returned: [No Error]

The delay_pool parameter for the daytime use is:
dealy_parameters 4 206000/206000 16000/51200. In a test I I doubled it to
32000/91200 and it still failed.
read_timeout is 15minutes (enough time for the ColdFusion job to finsih) but
still it fails.

The error 504 says Gateway Timeout but the proxy goes directly to the
internet. Squid is 2.5STABLE2, Server is an Ultra 60 running Solaris 2.8.
If the problem is in fact delay pools, is there some other parameter in the
squid.conf file that I can tweak or will I have to create an exclusive rule
giving this one site/project a larger slice of bandwidth? I already tried
that (doubling it) and still it fails. I checked the conf file but wasn't
sure what the source of the problem is (what Gateway is Timing out, do I
need to bump up one of the 4 or 5 tunable timeouts in the conf file,
or....).

Anyhow if you have any ideas, I would be most greatful. thanks,

Adam
Received on Fri Jun 13 2003 - 19:24:14 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 17:17:22 MST