Hi,

today I ran into some troubles with stream clients connected trough proxy. We are yet implementing CC into our streaming system and are so redirecting streams from our old server to the new one. (and at the same time we are trying to use proxy feature of CC3 wich would be a very nice addon

So, up to a certain amount of traffic, everything worked fine, but after reaching aprox. 10% of 1Gbit link the cc web access stoped to respond. Also all streams that used proxy server were disconnected.

Than I made a restart of cc service trough CLI and everything started working fine again - but again after reaching aprox. the same amount of trafic - same problem - disconnect of clients over proxy and no response from web console - again after restart ok  - and after reaching same amount of traffic problem again.
Then I redirected a few of streams back to our old servers so I did not reach this limit again. Up unitil now everything ok. 

Sever is on 1Gbit link dedicated to only this VM, Vm runs on VMware ESXi with CentOS, CC 3.1.1,
When this happend aprox. 700 128kbps connections ran over proxy ...

In the web access log:

First I streted getting this enteries:
2014/12/09 09:52:08 [error] 26290#0: *1 upstream sent no valid HTTP/1.0 header while reading response header from upstream, client:

 then after some time enteriues like this:
2014/12/09 10:31:16 [alert] 26290#0: *3250 1024 worker_connections are not enough while connecting to upstream, client:

and then also this:
2014/12/09 10:33:23 [alert] 26290#0: 1024 worker_connections are not enough

What could be the cause? Is proxy having some missconfiguration? What are the limits of using proxy? Is WMware the pbm?

BTW: I also notice that statistics for some streams are being incorect for the past 2 days...   dont know if this is related?

Thank you very much for your help.
David

Hi

I v done some additional investigation regarding statistics. First let me correct myself - it did not stop to work - its just that the numbers dropped drasticaly. - And as it looks the listener statistics numbers droped right after we implemented links trough proxy. Could this be the cause?

We are using Shoutcast DNAS 1 server only.

Does the listeners that are connected trough proxy not beeing statisticaly processed? (It makes sense beacuse as far as shoutcast log file is concerned they are all listed in as 1 unique listener.)?

It would be very usefull if you added option in statistics to show also the listeners that are connected trough proxy as a seperate category and summed all together for the correct grand total.

Regarding proxy I'm still in blank -

Could this nginx optimizations help in peformance?
https://www.digitalocean.com/community/tutorials/how-to-optimize-nginx-configuration

Thank you for your help and br, David

Just gotta do exactly what it says in the error message, increase worker_connections in (I think) /usr/local/centovacast/etc/cc-web.conf to how ever many simultaneous listeners you want to handle thru the proxy. Nginx is just putting the brakes on cuz if you're running a smaller server or vps, going over 1024 conns will probably make your server have a shit.

Proxy listeners are already in the stats so no change needed there... works fine on my servers.

Thank you very much for your reply.

Yes it seems that this should do the trick - but as I opened this config file I saw notice in the header that changing it would mean breach of contract with Centova.
Besides that also I dont know what such costumization would mean for future automatic updates of Centova... would it thsi config file be overwriten?

I gues the most perfect solution would be if centova added proxy fine tuning  settings also into the Centova web managment console? (At least worker processes  and connections?)
This way you could change them from the interface - and everything would also stay that way after future updates?

Thanks