no mrouted on multi-homed SGI O2
Craig M. Votava
I continue to be frustrated at trying to get
any version of mrouted to run correctly on a
multi-homed SGI O2.
Has anyone successfully done this at any time?
Please respond to my email address, as I'll be
on vacation until the end of October, and won't
be getting any MBONE mailing list email.
Attached are the current details. Please take
a look at them and help if you can.
Thanks
-Craig Votava
Lucent Technologies
SETUP:
======
NETWORK A | NETWORK B
A1 -------- AB -------- B1
A1 - A client on the A network, sourcing multicast audio and video
AB - The multi-homed SGI O2, running mrouted 3.9b3
B1 - A client on the B network trying to receive the source from A1
-All routers on both networks have mutlicasting turned off.
-AB has all of the lines in /etc/mrouted.conf commented out.
-AB is setup for the primary interface on network A
SYMPTOMS:
=========
-When A1 announces an sdr session with audio and video,
the session shows up in the sdr windows on A1,
AB, and B1.
-When B1 joins the session, it gets no multicast packets.
-When AB joins the session, AB receives video and audio,
and suddenly B1 starts getting audio and video too.
-When AB quits out of the video and/or audio tool B1 stops
getting the video and/or audio at the same time.
-When using mtrace to try and trace between A1 and B1, in
either direction, sometimes mtrace has to go to
point-to-point mode, other times it doesn't. This
seems to vary over time.
-In the "netstat -M" output, the multicast forwarding cache
has an entry for B1 on the audio and video multicast
group addresses. When B1 doesn't receive media, the
count numbers don't increase, when B1 does get media
the count numbers do increase.
-Running "netsnoop" on the A interface of AB shows continuous
multicast media packets at all times when A1 is
transmitting. This was verified by an independent PC
snooping on that same segment as well.
-Running "netsnoop" on the B interface of AB when B1 is not getting
media shows only the person ID packets (or whatever they're
called) on the media sessions. When I have B1 send continuous
audio packets, netsnoop on AB picks them up fine.
Angel Mateo
I think your problem is not you have a multi-homed SGI O2. I have one in just
one network and I have the same problem.
My machine has some tunnels with other networks and that works right, but in
my network, the other computers can only receive the sessions that receive the
O2 (the mrouter). If the O2 joins to a session, then the other computers can
receive media from this session, but otherway they can´t.
(Please, if you fix the problem, advice me what can I do to fix it)
-----------------------------------------
Angel L. Mateo Martinez
RedIRIS/CSIC
C/ Serrano, 142
28006 Madrid (Spain)
Tlfo: +34-91-5855145
Fax: +34-91-5855146
Thomas M. Kendall <tkendall>
I filed a bug report with SGI on this issue in March. I worked with
them for 3 months before they admitted that there is a problem in the O2
kernel multicasting code. They have now stated that their Engineering
Department has no interest in fixing it.
The SGI case number is 0917087.
-------------------------------------------------------------------------------
Thomas M. Kendall U. S. Army Research Laboratory
410-278-9195 Corporate Information and Computing Center
410-278-9199 (FAX) Attn: AMSRL-CI-HC
tken...@arl.mil Aberdeen Proving Ground, MD 21005-5067
-------------------------------------------------------------------------------
Ignatios Souvatzis
> FYI-
>
> I filed a bug report with SGI on this issue in March. I worked with
> them for 3 months before they admitted that there is a problem in the O2
> kernel multicasting code. They have now stated that their Engineering
> Department has no interest in fixing it.
>
> The SGI case number is 0917087.
Ah, that explains much....
A colleague in the neighbour department complained that he couldn't start
MBONE tools on his machine because it would instantly panic.
-is
Craig M. Votava
Tom, thanks for the heads-up on your sgi case number 0917087.
Its discouraging to hear that SGI Engineering has no
plans to fix this problem, however I believe that if
enough people bug them, there's a chance that they
may do something.
Following that logic, I encourage everyone who can, to
log a case with SGI on this topic. My case number is
1022105, and the text appears below. I like to think
that SGI is responsive to its customers (they should
take lessons from CISCO), but I've not had good
experiences with them.
All help here is appreciated!
Thanks
-Craig
================= SGI Case #1022105 ==================
When using the /usr/etc/mrouted command on an SGI O2
running any version of the IRIX operating system, multicast
packets are not distributed correctly.
I have been informed that this is due to a bug in the
IRIX kernel's multicasting code, and has been known
about since March (see SGI case number 0917087).
This situation as it exists is unacceptable.
Please let me know when and where a patch for this will
be made available so we can bring back up our multicasting
network.
If you cannot supply a fix by November 1, 1998 please
contact me by phone immediately so we can discuss this
matter.
Thanks
Craig Votava
Lucent Technologies
Amancio Hasty
towards native ip multicast support in your lan via a router which
supports native ip multicasting.
Regards,
Amancio