On Tue, 2007-05-22 at 18:07 +0200, Emmanuel Eyer wrote:
> Dear Squid team,
>
> I am using Squid 3.0 PRE6 in production (30 GB data and 2 M requests per
> day on 4 squid instances) and works pretty well. (BTW, it runs much more
> stable than older versions!)
>
> However, I have been reported that https connections break down after 15
> minutes, for example during large downloads (ISO files) or during video
> conferencing. According to some users, it is exactly 15 minutes (900
> seconds), systematically. Partial tests, bypassing the proxies, confirm
> that the proxy servers are culprit.
>
> In addition (though I do not know yet if it is linked to above issue), I
> found lots of entries in my cache.log files similar to:
> 2007/05/22 17:09:18| sslReadServer: FD 78: read failure: (0) Error 0
> (Lots meaning 50-100 per day and per instance. No such entries were
> found prior upgrading to PRE6, though disconnections were observed.)
>
> I checked through the configuration, but could not find any 900 seconds
> timeout. Notice that no such problem occurs with http connections.
>
> Quick configuration facts: Solaris 9 (SPARC), gcc build, no caching
> (cache_dir null), no SSL support (I do not use authentication).
>
> Could this be linked to bug 1633? Shall I comment on this bug? Or open a
> new one? I can setup a test server and turn on various debug options if
> you need, as well as providing more detailed information on my current
> Squid configuration.
Sounds like bug #1633 to me, but I do not know much about SSL-related
code. Anybody in-the-know care to comment?
Which version did you upgrade to PRE6 from? There were no 15-minute
breakages with that version, right?
Were you able to reproduce this bug yourself?
Thank you,
Alex.
Received on Thu May 24 2007 - 08:39:09 MDT
This archive was generated by hypermail pre-2.1.9 : Fri Jun 01 2007 - 12:00:09 MDT