Icecast Streaming Media Server

iunderwood · Joined: 23 Aug 2008 Posts: 114 Location: Leicester, MA

I've found a couple of caveats with the on-demand relay system when using a master/slave configuration.

* The slave mirrors all mountpoints, which is fine, except it's also shows the hidden mountpoints I have. Can hidden mounts be excluded from the streamlist?

* I have an update interval of 60 set in the slave config, but it only holds onto the last metadata received in a stream.

* Since I'm using a single master, single slave, I'm allowing listener connections to either the master or the slave. However, it's not necessarily easy to get an accurate listener count with an on-demad relay. Would it be possible to add a separate counter for relays and keep it separate from the listener count?

[Edit: Made some changes since I'm understanding the behavior better.]
_________________
++I;

karlH · Code Warrior Joined: 13 Jun 2005 Posts: 5476 Location: UK

If the hidden relay is being listed on the streamlist then it's a bug and we would need more details on the situation, version, log etc.

With an inactive relay, the metadata is not updated because it is in-stream, which is not being transferred.

I'm not sure what the issue is on the 3rd query. The listener count should be accurate for that individual mount.

karl.

iunderwood · Joined: 23 Aug 2008 Posts: 114 Location: Leicester, MA

The version in question is KH25.

Part 1: Hidden Relay in /admin/streams:

The URL call to /admin/streams is listing 6 streams, two of which I have hidden in the configuration.

karlH · Code Warrior Joined: 13 Jun 2005 Posts: 5476 Location: UK

yep the hidden setting was ignored, the wrong flag is being checked. A simple change there means it works as it should, kh26 should fix that.

Looks like /admin/streams was getting auth bypassed if there was no mount/auth defined (useful if you want url/htpasswd auth for slaves), fixed should be in kh26

As for the listener counts, the master does not collect the different listener counts yet. There is partial work to retrieve details like that but it isn't complete yet. I'm hoping to get the master to connect to the slave to retrieve the stats in an non-polling way, so it will be near-accurate.

I should be releasing kh26 shortly, there was a non-ogg metadata bug possible.

karl.

iunderwood · Joined: 23 Aug 2008 Posts: 114 Location: Leicester, MA

Much thanks for the fixes. I'm looking forward to making the latest compile.

On the listener vs. slave issue, I know there was a discussion a few years ago on using a master redirect. (http://icecast.imux.net/viewtopic.php?t=1440) That was back in the days of 2.3.1, but the thread is hanging and didn't indicate if the redirect worked or not. I didn't know the feature existed until yesterday, which got me all kinds of excited.
_________________
++I;

karlH · Code Warrior Joined: 13 Jun 2005 Posts: 5476 Location: UK

if you use a tag group on the slave like

<master>
<server>....</server> <port>...</port>
<redirect>true</redirect>
<username>...</username>
<password>...</password>
</master>

and optionally <max-redirect-slaves>10</max..> on the master

then the master will 302 new listeners when it is full. At the moment there is no feedback from the slave.

kh26 is up on my site now.

karl.

iunderwood · Joined: 23 Aug 2008 Posts: 114 Location: Leicester, MA

Upgraded to KH26, and the web interface is returning a lot of 403s:

96.32.118.18 - - [17/Aug/2010:09:40:14 +0000] "GET / HTTP/1.1" 403 51 "-" "Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10_6_4; en-us) AppleWebKit/533.17.8 (KHTML, like Gecko) Version/5.0.1 Safari/533.17.8" 0

I've backed to KH25 in the meantime.
_________________
++I;

karlH · Code Warrior Joined: 13 Jun 2005 Posts: 5476 Location: UK

looks like my fix for the admin streams thing was wrong, serves me right for doing changes right at the last moment, checking...

karl.

iunderwood · Joined: 23 Aug 2008 Posts: 114 Location: Leicester, MA

I'm playing with the master/slave thing as well, and would it be possible to put in the Manage Relays tab which would show known slaves?
_________________
++I;

iunderwood · Joined: 23 Aug 2008 Posts: 114 Location: Leicester, MA

And another strange bug while we're on a roll.

I've tried to set max_listeners for a certain mount to 0, to force it to redirect:

karlH · Code Warrior Joined: 13 Jun 2005 Posts: 5476 Location: UK

that one isn't my bug, you forgot the / in the mountpoint

I've re-uploaded the kh26 tarball if you want to try it again.

Yes those are the defaults, before the parse_xml_tags call

karl

iunderwood · Joined: 23 Aug 2008 Posts: 114 Location: Leicester, MA

D-oh! Trailing slash ... ugh.

I'll retry the tarball. Smile

EDIT:

Adding the / fixed the mountpoint. (Okay ... I'm up waay too late as it is). With a max of 0, it redirects after reset as soon as the next slave ping comes in. I know the LB is simplistic, but if there are multiple relays, how does it choose who goes where?

Hidden mounts are no longer listed to the slaves.

Streams are authenticated! (Restarted my browser to clear the PW in cache)
_________________
++I;

karlH · Code Warrior Joined: 13 Jun 2005 Posts: 5476 Location: UK

it's based on random selection at the moment. When feedback from the slave occurs then free slots can be determined. Eventually the /admin/streams polling will go as well in favour of the stats connection as that will report relays immediately and allow other details like yp passwords to migrate.

karl.

iunderwood · Joined: 23 Aug 2008 Posts: 114 Location: Leicester, MA

And another ODR oddity ... the YP call is updating for all mountpoints, based on last known metadata, which is often stale:

iunderwood · Joined: 23 Aug 2008 Posts: 114 Location: Leicester, MA

My next question with this setup is with URL authentication.

Does the master attempt authentication before the redirect? If so, is any authentication passed to the slave, or does the slave need to be separately configured to authenticate? I believe I saw an icecast-relay: 1 response which can be used to bypass the redirect, correct?
_________________
++I;