sc_trans bug report

[PieterPh]

Hi guys,

I'm in the process of setting up a radio station, and I found a bug you're probably interested in :-)

The setup is simple: I have one server that will record a live stream, encode it, and send it in two streams (40kbit and 128kbit) to two different servers (so four streams in total). Anyway, the configuration seems to be working perfectly, but I found out it still had one major problem: after a while, the stream would suddenly stop for about 30 seconds, after which it got disconnected. The weird thing is that I never noticed it the previous week, so it appeared to have come out of nothing.

I decided to investigate the issue, and found out that the stream stopped simply because sc_serv didn't have more data to send. For some reason, the flow of data from the other server (running sc_trans) simply stopped. After 30 seconds, the 'autodumpsourcetime' timeout would kick in and dump the source stream. At the same moment a different timeout would kick the user (because sc_serv hadn't sent data for 30s). Now, the interesting thing is that when sc_serv closed the connection, sc_trans would reconnect and everything would work again.. for a little while.

I looked through the log files of sc_trans, and the only thing that I could find was that at some point it would log an error where it says it could not send bytes to sc_serv.

In order to further investigate the issue, I decided to simplify my config file. This basically means that I commented two of the four streams. Interestingly, the issue suddenly disappeared and everything was working perfectly again. And that's when it hit me: the server I commented out wasn't running when I was having this problem! So apparently sc_trans doesn't do a good job of handling a server that is down.

To make sure that this was indeed the source of the problem, I restored my config file and set up two instances of sc_serv. Everything was working perfectly without a single problem for over 10 minutes. Then I killed one of the sc_serv instances (the one I wasn't listening to), and sure enough... after about a minute WinAMP was stalling again and the problem reappeared.

So, in summary: it seems that when sc_trans is configured to send data to a server that is not running, it eventually blocks all network streams including the streams to the other (active) servers. Normally not a big problem when all the servers are running properly, but if one of the sc_serv servers crashes, it might kill the entire station!

Anyway, I hope you can reproduce and fix the issue (I'm running the 32-bit Windows version of sc_trans)...

Regards,
Pieter

[DrO]

something close to this has already been reported in the main sc_trans discussion thread and is logged to be investigated since any loss of connection be it a sc_serv killed or the internet connection failing. am surprised it never cropped up in the previous sc_trans drop 6 thread (at least i don't remember seeing it mentioned when i went through all of the posts so could be something i've introduced from taking up dev on the tools).

-daz