View previous topic :: View next topic |
Author |
Message |
karlH Code Warrior
Joined: 13 Jun 2005 Posts: 5476 Location: UK
|
Posted: Sat Jan 07, 2012 2:53 am Post subject: |
|
|
can you try www.xiphicecast.webspace.virginmedia.com/icecast-2.3.2-kh31.2_setup.exe.
Seeing that I've no comments from anyone else about this, I decided to look at win32 specific calls. It could be down to something not quite clear in the docs for win32. If it is this then the behaviour would be shown unrelated to relays starting, but instead, listeners would experience sync errors. This would only affect FLV listeners currently so if you have none then it won't be down to that.
karl. |
|
Back to top |
|
|
robertut
Joined: 31 Aug 2007 Posts: 156
|
Posted: Wed Jan 18, 2012 9:04 pm Post subject: |
|
|
Thanks, tested.
Same as previous, unfortnately
|
|
Back to top |
|
|
karlH Code Warrior
Joined: 13 Jun 2005 Posts: 5476 Location: UK
|
Posted: Wed Jan 18, 2012 10:45 pm Post subject: |
|
|
A bit of a head scratcher, debugging based on only a graph to go on is a bit stretch. Again no reports elsewhere so I really think this is down to some win32 specific issue. One other possibility is that the OS is reporting a full use of it's buffers and that is translating to an unrecoverable error dropping the listener.
Assuming this particular case is recoverable (just heavy use of the bursting with lots of new listeners) and just requires the listener to back off to allow data to go, then maybe the next build will help.
http://www.xiphicecast.webspace.virginmedia.com/icecast-2.3.2-kh31.4_setup.exe
karl. |
|
Back to top |
|
|
robertut
Joined: 31 Aug 2007 Posts: 156
|
Posted: Thu Jan 19, 2012 11:39 am Post subject: |
|
|
Thanks, I'll give it a shot.
Next question: is there a way to do geo-blocking in Icecast server?
It's a mandatory requirement for the Olympic games... |
|
Back to top |
|
|
karlH Code Warrior
Joined: 13 Jun 2005 Posts: 5476 Location: UK
|
Posted: Thu Jan 19, 2012 12:07 pm Post subject: |
|
|
You can use the listener_add auth url trigger, that can use the IP as as reject or accept condition. I'm sure some of the geoip utils have say a php interface where IP lookup can be used.
karl. |
|
Back to top |
|
|
level20peon
Joined: 24 Feb 2008 Posts: 19
|
Posted: Tue Feb 07, 2012 10:38 am Post subject: |
|
|
Anonymous wrote: |
If you have a look at my config a page before, moving the:
Code: |
<shoutcast-mount>/stream.nsv</shoutcast-mount> |
part above the listening socket definitions did the trick. |
The example "icecast.xml" configuration in the GIT trunk was not updated with this change, yet.
It is possible to either include the <shoutcast-mount> tag above all <listen-socket> tags, or instead, enter it like proposed in the "icecast_shoutcast_compat.xml". |
|
Back to top |
|
|
karlH Code Warrior
Joined: 13 Jun 2005 Posts: 5476 Location: UK
|
Posted: Tue Feb 07, 2012 5:44 pm Post subject: |
|
|
either is allowed, it is just that the per listen-socket way is more flexible. The global method is mainly to allow for the 2.3.2 way of using a shoutcast-compat option.
karl. |
|
Back to top |
|
|
robertut
Joined: 31 Aug 2007 Posts: 156
|
Posted: Wed Feb 08, 2012 10:47 pm Post subject: |
|
|
Karl, look at this:
This is good. That's what we get with KH22. More than 5000 listeners on the 25 different streams, at 450 Mbit/s served from icecast Win32 service.
Recent builds can't seem to be able to handle that. Except KH30 which was nearly the same as this but it had relay reconnection problems. |
|
Back to top |
|
|
karlH Code Warrior
Joined: 13 Jun 2005 Posts: 5476 Location: UK
|
Posted: Thu Feb 09, 2012 1:40 am Post subject: |
|
|
Have you been able to run into this with simulated listeners yet. Have you determined if the issue is down to sources stopping or just a certain number of listeners dropping, if listeners were dropped then was it because they were slow?
Running a lesser load test here had no results so far. The more recent fixes over the last few weeks wouldn't just affect some listeners across various mountpoints.
karl. |
|
Back to top |
|
|
robertut
Joined: 31 Aug 2007 Posts: 156
|
Posted: Mon Feb 13, 2012 3:08 pm Post subject: |
|
|
Nope. This appears only with high number of listeners. It's our main (production) server, with full national and international listening audience.
Have no clue how to simulate this amount of listeners in a test-environment. |
|
Back to top |
|
|
karlH Code Warrior
Joined: 13 Jun 2005 Posts: 5476 Location: UK
|
Posted: Mon Feb 13, 2012 3:20 pm Post subject: |
|
|
usually best to run the listeners on a separate box (maybe near by but on a decent connection link). Then run something like ab or a script with multiple curls to extract a sample from the stream, could even be rate limited.
You don't need to stop the production setup, just run the new binary on a separate port, maybe relaying the from the production server. The test is for load levels not for playback.
karl. |
|
Back to top |
|
|
robertut
Joined: 31 Aug 2007 Posts: 156
|
Posted: Wed Feb 15, 2012 10:23 am Post subject: |
|
|
I don't think I can do that now...
Meanwhile: since the downgrade to KH22 (beginning of week 5), we're able to serve 25% more listeners than before:
|
|
Back to top |
|
|
aoz1
Joined: 05 Sep 2010 Posts: 6 Location: UA
|
Posted: Mon Feb 20, 2012 12:54 am Post subject: |
|
|
Karl, I just try to play HD video stream over icecast.
you somehow limit the limit_rate up to 2Mbits/sec.
Ho do I rise it up to 4Mbits/sec.
Thanks in advance |
|
Back to top |
|
|
karlH Code Warrior
Joined: 13 Jun 2005 Posts: 5476 Location: UK
|
Posted: Mon Feb 20, 2012 1:40 am Post subject: |
|
|
it's isn't a figure that is imposed directly that can be lifted, it's probably a combination that I've not specifically checked for, after all it's not a common bitrate of stream
I'll try a few tests to make sure, what figure are you actually specifying in limit-rate?
karl. |
|
Back to top |
|
|
aoz1
Joined: 05 Sep 2010 Posts: 6 Location: UA
|
Posted: Mon Feb 20, 2012 10:11 am Post subject: |
|
|
karlH wrote: |
it's isn't a figure that is imposed directly that can be lifted, it's probably a combination that I've not specifically checked for, after all it's not a common bitrate of stream
I'll try a few tests to make sure, what figure are you actually specifying in limit-rate?
karl. |
It's not absolutelely necessary to use the limit-rate. My version have already a lot of of hack-s. So, I can just reassign some set of constants in source. |
|
Back to top |
|
|
|