Le 11/04/2012 20:37, Jean-Philippe Menil a écrit :
> On 11/04/2012 14:49, Amos Jeffries wrote:
>> On 11/04/2012 6:47 p.m., Jean-Philippe Menil wrote:
>>> Le 05/04/2012 16:50, Jean-Philippe Menil a écrit :
>>>> Le 04/04/2012 09:00, Jean-Philippe Menil a écrit :
>>>>> On 03/04/2012 23:53, Amos Jeffries wrote:
>>>>>> On 04.04.2012 02:46, Jean-Philippe Menil wrote:
>>>>>>> Le 03/04/2012 11:06, Jean-Philippe Menil a écrit :
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> i encounter serious outage with squid 3.HEAD-20120307-r12077.
>>>>>>>> Every time i download some test files, it stop after 15 minutes.
>>>>>>>> If i go down read_timeout to 1 minutes, the download stop after 1
>>>>>>>> minutes.
>>>>>>>>
>>>>>>>> Is it a know issue, or must i increment read_timeout to
>>>>>>>> excessively
>>>>>>>> timeout?
>>>>>>>>
>>>>>>>> special configuration is as follow:
>>>>>>>>
>>>>>>>> workers 4
>>>>>>>> cpu_affinity_map process_numbers=1,2,3,4 cores=6,7,8,9
>>>>>>>>
>>>>>>>> Regards.
>>>>>>>>
>>>>>>> Nobody have ever observe this phenomen?
>>>>>>
>>>>>> Not many production networks (squid-users people) use 3.HEAD
>>>>>> (alpha) code.
>>>>>> The developers and alpha/beta testers hang out in squid-dev ;)
>>>>>>
>>>>>> And no, you are the first to mention this particular behaviour.
>>>>>>
>>>>>> Amos
>>>>>>
>>>>> Hi,
>>>>>
>>>>> yes iknow, but i think it is present in 3.2 too (i will test this
>>>>> afternoon to confirm).
>>>>> I think i can repeat that only when download a file to on a https
>>>>> site, does it help?
>>>>>
>>>>> Regards.
>>>>>
>>>> Hi,
>>>>
>>>> so i have done test with a squid 3.2.0.14.
>>>> And it appear that i can repeat the problem only with https site, why
>>>> i don't know yet.
>>>>
>>>> For test, i fixe a lower value (don't want to wait 15 minutes between
>>>> each test) for read_timeout,
>>>> and i download an iso file through some https site:
>>>> https://nzdis.org/projects/projects/perfnet/repository/revisions/4/raw/vendor/Vyatta/Vyatta/vyatta-livecd-vc5.0.2.iso
>>>>
>>>>
>>>>
>>>> Every time, the download stop at the read_timeout value.
>>>>
>>>> Any ideas?
>>>>
>>>> Regards.
>>>>
>>>>
>>>
>>> Hi,
>>>
>>> sorry to up this subject, but i can't understand why the
>>> read_timeout isn't zeroed with https communication.
>>> Do i miss something?
>>
>
>> Read timeout should be reset to full on every packet read. It is never
>> zeroed during a transfer.
>>
> My fault, bad semantic language, i mean reset, effectively.
>
>> It sounds like the transfer is stalling for more than read timeout, or
>> the timeout is not being reset like it should be.
>>
> Do you know, how i can troubleshoot this issue?
>
>> Amos
>
> Many thanks.
> Regards
>
Hi Amos,
I think i got it:
i still encounter the problem with 3.2.0.17.
I can easily repeat with the following parameter
read_timeout 30 seconds
wget --no-check-certificate
https://nzdis.org/projects/projects/perfnet/repository/revisions/4/raw/vendor/Vyatta/Vyatta/vyatta-livecd-vc5.0.2.iso
-O /dev/null
The donwload will stop on the fixed read_timeout.
With the following (obvious) patch, the download don't stop anymore:
01-build64:/home/menil-jp/squid-3.2.0.17# diff src/tunnel.cc
/tmp/squid-3.2.0.17/src/tunnel.cc
321,326d320
< if (Comm::IsConnOpen(to.conn)) {
< AsyncCall::Pointer timeoutCall = commCbCall(5, 4, "tunnelTimeout",
< CommTimeoutCbPtrFun(tunnelTimeout, this));
< commSetConnTimeout(to.conn, Config.Timeout.read, timeoutCall);
< }
<
333,336c327,330
< /* Only close the remote end if we've finished queueing data to it */
< if (from.len == 0 && Comm::IsConnOpen(to.conn) ) {
< to.conn->close();
< }
--- > /* Only close the remote end if we've finished queueing data to it */ > if (from.len == 0 && Comm::IsConnOpen(to.conn) ) { > to.conn->close(); > } Can you confirm the problem? Best regards. -- Jean-Philippe Menil - Pôle réseau Service IRTS DSI Université de Nantes jean-philippe.menil_at_univ-nantes.fr Tel : 02.53.48.49.27 - Fax : 02.53.48.49.09Received on Tue Apr 17 2012 - 12:06:26 MDT
This archive was generated by hypermail 2.2.0 : Tue Apr 17 2012 - 12:00:03 MDT