[SIP Beyond VoIP] MSRP relay very unstable
prasun.bheri at gmail.com
Mon Apr 21 13:27:55 CEST 2014
Issue seems to be linked with large number of chunks in a transfer.
We have tested our setup with Blink client (latest version 0.8.0) and
noticed that client is chunking file at around 60k bytes per chunk. at this
rate there are far fewer chunks compared to our scenario where chunking is
done for every 2k bytes.
When we configure our client to chunk at 60k its behavior is much similar
to Blink; majority of file transfers are succeeding however there are still
some failures similar to Blink client (on transferring 30 files [~5mb each]
simultaneously between two Blink clients 2 got cancelled automatically).
To identify data corruption I have generated a file containing numbers in
increasing order and transferred this file over tcp. In wireshark we can
clearly see that file is sent to server successfully however at receiving
end additional bytes (portion of 7 hyphens transaction id built by source
client) are present. Please have a look at these wireshark captures
uploaded here http://msrprelay.org/issues/2501.
If you watch the tcp stream of data transfer in receivers capture note that
"-----" is sent extra after number 58682. similarly after number 115492
"-------YS7dPeoD+\r" is sent extra and lastly after number 270628
"-------O" is sent extra.
Thanks & Regards,
On Tue, Apr 15, 2014 at 1:48 PM, <ag at ag-projects.com> wrote:
> Well, MSPRelay is used for several years in production and have not heard
> such coruption problems. I personably transfer lot of large files with it
> and I never ended up with a corrupted file. Also I use it for chat and
> screen sharing and always works for me.
> I use Blink as a client. Maybe it has something to do with your client.
> On 15 Apr 2014, at 08:55, Prasun Bheri <prasun.bheri at gmail.com> wrote:
> Hi Group,
> I have configured msrp relay from http://msrprelay.org with plain tcp and
> did some file transfers.
> I find it very unstable as transfer randomly gets aborted midway (relay
> sends reset to both clients) when doing a 2mb file transfer with small (2k)
> chunk size. but having a better luck when chunk size is 10k or greater as
> long as there is single transfer happening. when transferring multiple
> files one or all the transfers are getting aborted by relay sending reset
> to clients.
> This looks like some buffer issues.
> has anyone faced this or similar problems? if any one know a fix for this
> please let me know.
> These are the configuration options I am using:
> address = 172.31.5.219:2855
> debug_notls = yes
> backend = memory
> 55575 at 18.104.22.168 = 575.55 ### using public ip
> 55576 at 22.214.171.124 = 576.55
> moreover at times, data transferred is getting corrupted. I mean received
> data randomly contains portion of "7 hyphens <transaction id>" sequence
> sent by sending client. there must be some issue with re-packetization of
> chunks being forwarded. I raised this as defect here
> ps: does any one know any other msrp relays servers paid/free that are
> stable and time tested.
> Thanks & Regards,
> SIPBeyondVoIP mailing list
> SIPBeyondVoIP at lists.ag-projects.com
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the SIPBeyondVoIP