[Freeswitch-users] mod_httapi issues
colin.morelli at gmail.com
Tue Jul 12 19:00:55 MSD 2016
Yeah, the content-length appears to be correct. I didn't look at the delay
between the packets, but I can check on that (is there a low default read
timeout in httapi?)
What ends up happening is that FS receives a partial XML response,
processes all of the complete instructions, and then re-requests the same
URL again. Sometimes on the second time it works. In the worst case, I have
seen FS loop over the request 6 times before eventually it receives a full
Also, for what it's worth, I build FS off of master for a different issue I
was trying to fix, and consequently ran it outside of docker (on Ubuntu
Xenial) and the same issue happened. So, that rules out anything specific
to that FS version and/or docker. Most likely curl issues also (unless
Xenial and the libcurl in the jessie container both had the bug).
I'll try to grab a pcap and send it your way.
On Tue, Jul 12, 2016 at 10:55 AM Anthony Minessale <
anthony.minessale at gmail.com> wrote:
> Is the content length on the response consistent with the bytes in both
> packets together?
> is there any delay between the 2 packets.
> I might have to see the pcap to really comment.
> On Mon, Jul 11, 2016 at 4:13 PM, Colin Morelli <colin.morelli at gmail.com>
>> Hey all,
>> I've previously been using mod_httapi with no problems - so I'm not quite
>> at the point of calling this a FS bug yet (although the behavior is
>> certainly unexpected).
>> I recently set up a t2.nano instance in AWS - which are 512m memory. The
>> purpose of this instance is just to test FS configurations. Generally
>> speaking it works fine, with one exception:
>> Relatively frequently (throwing out a number of 15% here which may not be
>> right but it's in the right ball park), the response that FS receives from
>> my endpoint is cut off at some point throughout the string. When it
>> happens, the last character that FS receives exactly matches the final
>> character of a single TCP packet. However, after running tcpdump on the
>> machine I do see the full response coming back (split across two packets).
>> Now, I had not seen this problem until launching the t2.nano instance.
>> Three things have changed since then:
>> 1. FS 1.6.7 -> 1.6.9 (though I don't think this matters, as I don't see
>> any relevant httapi changes)
>> 2. Moved into a docker container
>> 3. Reduction in instance specs (2 cores + 2GB ram down to 1 core + 512mb
>> I should note it's also possible that the dependencies (such as libcurl)
>> have changed, though I searched through libcurls bug list and couldn't find
>> anything related to this.
>> Because of item #3 my first assumption was a memory allocation issue,
>> though FS reports nothing about this in the logs and when it happens I can
>> see that the system still has 100+MB free.
>> I thought it would help to post here first to see if the general
>> consensus is "there's no way FS will run correctly on 512mb of ram" or
>> something along those lines. But please let me know if I should make a bug
>> for this.
>> Professional FreeSWITCH Consulting Services:
>> consulting at freeswitch.org
>> Official FreeSWITCH Sites
>> FreeSWITCH-users mailing list
>> FreeSWITCH-users at lists.freeswitch.org
> Anthony Minessale II ♬ @anthmfs ♬ @FreeSWITCH ♬
> ☞ http://freeswitch.org/ ☞ http://cluecon.com/ ☞
> ☞ irc.freenode.net #freeswitch ☞ *http://freeswitch.org/g+
> ClueCon Weekly Development Call
> ☎ sip:888 at conference.freeswitch.org ☎ +19193869900
> Professional FreeSWITCH Consulting Services:
> consulting at freeswitch.org
> Official FreeSWITCH Sites
> FreeSWITCH-users mailing list
> FreeSWITCH-users at lists.freeswitch.org
-------------- next part --------------
An HTML attachment was scrubbed...
Join us at ClueCon 2016 Aug 8-12, 2016
More information about the FreeSWITCH-users