[SR-Users] evapi tcp parsing??
Daniel-Constantin Mierla
miconda at gmail.com
Fri Sep 18 10:11:03 CEST 2015
Hello,
I just pushed a patch to master branch that should cope with partial
data received on tcp connection. No time to test at all, therefore any
feedback will be appreciated.
Cheers,
Daniel
On 15/09/15 14:52, Daniel-Constantin Mierla wrote:
> Hello,
>
> I will look if there are options in libev to buffer data or try to
> implement a buffering mechanism locally for such cases.
>
> Cheers,
> Daniel
>
> On 14/09/15 23:00, Jayesh Nambiar wrote:
>> Hello Daniel,
>> After further testing with evapi module, I figured that when
>> Netstrings are used, an event route is invoked individually for each
>> message even if if multiple netstring messages are received in a
>> single TCP packet. But this doesn't work effectively when a single
>> proper message is split-up in two packets. For Example, if a message
>> arrives as:
>> 12:Hello World!, 12:Hello World!, 12:Hello World! in a single packet,
>> kamailio properly invokes the event route "evapi:message-received"
>> thrice for every individual proper netstring message.
>> But if the first packet contains:
>> 12:Hello World!, 12:Hello
>> And Second Packet contains:
>> World!, 12:Hello World!
>> the event route is invoked only once !!
>>
>> The above pattern is very much possible while sending and receiving
>> packets over TCP Socket. Our tests for receiving an approximately 150
>> byte message over evapi socket at the rate of roughly 1000cps causes
>> a lot of real events to be missed because of the above problem. You
>> can never be sure when TCP will split messages in different chunks.
>> This definitely looks like a bug which makes it not very reliable at
>> large scale deployments. Would really appreciate your inputs on this.
>> Thanks;
>>
>> - Jayesh
>>
>>
>>
>> On Thu, Sep 10, 2015 at 4:01 PM, Jayesh Nambiar <jayesh1017 at gmail.com
>> <mailto:jayesh1017 at gmail.com>> wrote:
>>
>> Hi Daniel,
>> Thanks for the quick response. So if I do not use Netstrings,
>> does Kamailio allow me to create a custom logic in the script.
>> For eg. if I decide to use newline as a delimiter, can I keep
>> buffering the message until I encounter the delimiter from the
>> event route and then execute whatever I have to within the script??
>>
>> Thanks,
>>
>> - Jayesh
>>
>> On Thu, Sep 10, 2015 at 1:29 PM, Daniel-Constantin Mierla
>> <miconda at gmail.com> wrote:
>>
>> Hello,
>>
>> tcb is stream protocol and several messages can be queued on
>> the pipe at the same time. That is the reason for netstring
>> format, to be able to easily detect the boundaries of each
>> message. If netstring format is enabled and kamailio receives
>> several messages at once, it splits them and for each is
>> executing the event route.
>>
>> If netstring format is not used, the kamailio is executing
>> the event route with the entire content that was read at once
>> from the tcp connection.
>>
>> Cheers,
>> Daniel
>>
>>
>> On 09/09/15 22:01, Jayesh Nambiar wrote:
>>> Hello,
>>> I'm exploring the evapi module for my kamailio to interface
>>> with an external node.js app for third party stuff like AAA,
>>> billing engine tasks, notifications and so on. I followed
>>> and took some ideas from the rtjson and evapi tutorial found
>>> here(http://kb.asipto.com/kamailio:k43-async-sip-routing-nodejs)
>>> to build the node.js app consuming events.
>>> When I stress tested the scenario using SIPp and tried
>>> sending a lot of events at 300-350cps from Kamailio, I
>>> noticed that at times the client is receiving 2-3 events in
>>> a single message together although I do event_sync_relay
>>> once per SIP message received and have netstrings enabled. I
>>> believe this is a typical behavior of TCP and needs to be
>>> handled by the client using some kind of Netstring handler.
>>> Please correct me if I'm wrong.
>>> And hence I'd like to know what particularly needs to be
>>> taken care of while writing a client that is listening for
>>> events on raw tcp socket and how does kamailio handle this
>>> situation while receiving messages over TCP socket?? Does
>>> kamailio recognize the end of netstring properly on
>>> evapi:message-received and give exactly one message to take
>>> care of on every "message-received" event or should that be
>>> handled in the script somewhere !!
>>> I also referred cgrates client over evapi example which is
>>> written in GO, but I couldnt find them handling TCP streams
>>> clearly either.
>>> I'd really appreciate some expert suggestion here to make an
>>> informed decision on using the evapi module for a large
>>> scale solution.
>>>
>>> Thanks,
>>>
>>> - Jayesh
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> SIP Express Router (SER) and Kamailio (OpenSER) - sr-users mailing list
>>> sr-users at lists.sip-router.org
>>> <mailto:sr-users at lists.sip-router.org>
>>> http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-users
>>
>> --
>> Daniel-Constantin Mierla
>> http://twitter.com/#!/miconda <http://twitter.com/#%21/miconda> - http://www.linkedin.com/in/miconda
>> Book: SIP Routing With Kamailio - http://www.asipto.com
>> Kamailio Advanced Training, Sep 28-30, 2015, in Berlin - http://asipto.com/u/kat
>>
>>
>> _______________________________________________
>> SIP Express Router (SER) and Kamailio (OpenSER) - sr-users
>> mailing list
>> sr-users at lists.sip-router.org
>> <mailto:sr-users at lists.sip-router.org>
>> http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-users
>>
>>
>>
>
> --
> Daniel-Constantin Mierla
> http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
> Book: SIP Routing With Kamailio - http://www.asipto.com
> Kamailio Advanced Training, Sep 28-30, 2015, in Berlin - http://asipto.com/u/kat
--
Daniel-Constantin Mierla
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
Book: SIP Routing With Kamailio - http://www.asipto.com
Kamailio Advanced Training, Sep 28-30, 2015, in Berlin - http://asipto.com/u/kat
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sip-router.org/pipermail/sr-users/attachments/20150918/0e0f8f29/attachment.html>
More information about the sr-users
mailing list