[Serusers] Issues with SER for TCP load testing

Klaus Darilion klaus.mailinglists at pernau.at
Wed Nov 30 20:39:13 CET 2005


Stephen McVarnock wrote:
> Hi folks,
>    I am new to the mailing lists - unsure as to which list is the most 
> appropriate I have mailed to both - I apologise for this in advance!
> 
> I am using SIPP and SER(v 0.9.4) to load test an application - SIPP is 
> sending calls via UDP to SER at 15 cps and with a duration of 180 
> seconds. SER is then forwarding on the calls to the application using 
> TCP and receiving the replies back over TCP.
> 
> However I am seeing quite a few errors in SER after a while. I'd like to 
> determine if the issues I am seeing are with SER or with the application 
>  itself so I'd appreciate some help with the following errors.
> 
> Can someone tell me, for each of the error traces below,  what is the 
> root cause of it and (even better) how I can resolve it for SER - if it 
> is indeed an issue with SER and not with the application.
> 
> 1) 123(14715) ERROR: tcpconn_new: mem. allocation failure
>    123(14715) ERROR: tcp_main_loop: tcpconn_new failed, closing socket
> 

I guess this happens when ser runs out of shared memory.

For the other errors it might be usefull to use tcpdump to analyze the 
corresponding TCP connection.

regards
klaus

> 2) 54(14646) ERROR: tsend_stream: failed to send: (32) Broken pipe
>    54(14646) ERROR: tcp_send: failed to send
> 
> 3)  65(14785) ERROR: tcp_blocking_connect: SO_ERROR (146) 
> Connection      refused
>    65(14785) ERROR: tcpconn_connect: tcp_blocking_connect failed
>    65(14785) ERROR: tcp_send: connect failed
>    65(14785) msg_send: ERROR: tcp_send failed
> 
> 4) 63(14783) ERROR: tsend_stream: send timeout (10000)
>    63(14783) ERROR: tcp_send: failed to send
> 
> 5) There was also another error on the TCP select loop - select failed: 
> Too many connections (or something similar).
> 
> I have been playing around with the number of descriptors set using 
> limit but without success.
> 
> Like I said, my main concern is to eliminate SER where possible as the 
> cause of these issues before focusing attention on the application.
> 
> Thanks in advance,
>    Steve.
> 
> 
> 
> 
> 
> 
> 
> _______________________________________________
> Serusers mailing list
> serusers at lists.iptel.org
> http://lists.iptel.org/mailman/listinfo/serusers
> 
> 




More information about the sr-users mailing list