[squid-users] ICAP and Allow 204 Header
Gilles Bardouillet
gilles.bardouillet at atos.net
Mon Feb 8 17:01:49 UTC 2016
OK,
I located my issue in Adaptation::Icap::ModXact::canBackupEverything()
with the TheBackupLimit (64k)
I tried to change the value of SQUID_TCP_SO_RCVBUF by without success :-(
This value is set with BodyPipe::MaxCapacity;
How can I set this MaxCapacity ?
Regards,
Gilles.
Le 28/01/2016 17:46, Gilles Bardouillet a écrit :
> Sorry for the response form but I dont received the Alex email, so I
> tried below to recompose the thread discussion
>> On 01/25/2016 10:28 AM, Gilles Bardouillet wrote:
>>
>> >/I'm using SQUID with CAS ICAP Server but I have one issue : />//>/*
>> for some images, squid receive icap error as ICAP_ERR_OTHER /
>> It may be useful to know more details about that ICAP error. What ICAP
>> response, if any, does Squid receive when it generates ICAP_ERR_OTHER?
> Here is some details from debug mode :
>
> 2015/12/09 11:32:11.786 kid3| 93,5| ModXact.cc(653) parseMore: have
> 182 bytes to parse [FD 32;Rr/w job924]
> 2015/12/09 11:32:11.786 kid3| 93,5| ModXact.cc(654) parseMore:
> ICAP/1.0 200 OK
> X-Apparent-Data-Types: JPG
> Service: CAS 1.3.1.1(170722)
> Service-ID: avscanner
> ISTag: "56680096"
> Encapsulated: req-body=0
> Date: Wed, 09 Dec 2015 10:32:19 GMT
>
>
> 2015/12/09 11:32:11.786 kid3| 93,5| ModXact.cc(749) parseHeaders:
> parse ICAP headers
> 2015/12/09 11:32:11.786 kid3| 93,5| ModXact.cc(1079) parseHead: have
> 182 head bytes to parse; state: 0
> 2015/12/09 11:32:11.786 kid3| 93,5| ModXact.cc(1094) parseHead: parse
> success, consume 182 bytes, return true
> 2015/12/09 11:32:11.786 kid3| 93,3|
> ../../../src/base/AsyncJobCalls.h(177) dial:
> Adaptation::Icap::Xaction::noteCommRead threw exception: Invalid ICAP
> Response
> 2015/12/09 11:32:11.786 kid3| 93,4| Xaction.cc(514) setOutcome:
> ICAP_ERR_OTHER
>
> Do you need more ?
>>
>>
>> >/* I noticed that for all these errors, Squid dont send the HTTP
>> header />/Allows 204 /
>> Allow:204 is not an HTTP header field. It is an ICAP header field.
> Right
>>
>>
>> >/* I read the code and find the Allow 204 header _is only set when
>> />/preview is enabled_. /
>> Are you sure? Several factors affect ICAP Allow:204 request header
>> presence. Preview availability should not be one of them because
>> Allow:204 is about 204 responses _outside_ of Preview. See RFC 3507
>> Section 4.6.
> Right, preview is only used for Allow 204 In and not Out
> My case is about Allow 204 out.
>
>
> here is the source code from 3.5.13 fromModXact.cc:
> const bool allow204in = preview.enabled(); // TODO: add
> shouldAllow204in()
> const bool allow204out = state.allowedPostview204 = shouldAllow204();
> ....
> else if (allow204out)
> allowHeader = "Allow: 204\r\n";
>>
>>
>> >/My icap conf activated preview and preview size as follow :
>> />/icap_preview_enable on />/icap_preview_size 1024 /
>> IIRC, Squid ignores icap_preview_size in squid.conf (a bug). The ICAP
>> service OPTIONS response determines the Preview size (subject to an
>> internal limit of 64KB).
> My ICAP server (CAS) dont send any Preview size in OPTIONS response :-(
>> >/I read that the preview size value can be overwritten by OPTIONS
>> />/requests, so can give me some details, hints in order to find why
>> some />/pictures dont offer preview and then fails ? /
>> See RFC 3507 Section 4.5 for details on how Preview is negotiated. If
>> you think Squid violates the ICAP protocol, please file a bug report
>> with the corresponding capture of ICAP messages (from and to Squid).
>>
>> As for ICAP 204 outside of Preview, I believe Squid can offer to support
>> that ICAP response if all of the checks below are successful:
>>
>> * the origin server OPTIONS response includes Allow:204;
>> * the message content length is known at the ICAP request time; and
>> * the message content length does not exceed 64KB.
> Thanks, I will check theses things.
>> If you prefer to analyze the code, see
>> Adaptation::Icap::ModXact::shouldAllow204() and
>> Adaptation::Icap::ModXact::canBackupEverything().
>>
>>
>> HTH,
>>
>> Alex.
> Regards,
> Gilles.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squid-cache.org/pipermail/squid-users/attachments/20160208/abd5b5c8/attachment-0001.html>
More information about the squid-users
mailing list