Multicast Session on wrong eth-port
-
Hello,
We’ve noticed an issue with a fresh Trunk installation (and upgrade after that) that the multicast session gets started on the wrong eth port.
We’ve found an article on this, explaining that you can select a eth port in the storage nodes configuration but, apparently eth0 is selected there, but the multicast service outputs to eth1. This is a nic that is not connected.
We are running fog on ESXi, with 2 nics attached to the VM
-
@Tom-Elliott Yes, it is working now.
Thanks,
-
@dvchuyen Please update and see if all is a little better. I found I missed a grep statement in looking for the interface and hopefully have a correction for this.
Thanks for the info thus far.
-
@Tom-Elliott Please check my last post on this thread
-
I just updated to lasted Trunk version today and have the same problem
Multicast setting
Storage node (I have only 1 node)
And this is multicast log (wrong interface and IP address)
-
Any way we can get some more information with this?
-
As far as I am concerned: Yes, but the issue itself might not yet be fixed
-
Should we mark this solved for now as taking the second NIC down fixed the issue?
-
Hi Tom, sorry for me not replying. I was busy with different work in the meantime.
I am unable to recall the IPadress of the NIC, because the chosen NIC was connected to a VLAN that does not have a DHCP-server nor a static IP assigned. Also; as soon as we took the other NIC down, the problem resolved itself. I am willing to check what happens if I re-enable the second NIC, but am not able to do so in the upcoming week.
-
Can we get some info when you have time?
-
@abos_systemax When the command starts, I know it’s showing on the wrong interface, but is it also showing the wrong IP address?
-
@Tom-Elliott Trunk version 7843.
-
What version of FOG are you running?
I ask because, @Sebastian-Roth is correct that I changed ethernet discovery for multicast tasks.
Ultimately, the new code takes the Storage Node’s IP/Hostname and tries to find the interface that’s using that IP Address. (I convert hostname to IP address in case it’s stored on the Node in FQDN/Hostname form.)
How well it works, I don’t know as I can not test all potential scenarios.
-
@abos_systemax If you’re not using eth1, I think you can disable it in Linux for the time being, but this is definitely a bug.
-
all settings were set to eth0. As soon as we removed eth1, the multicast got onto the right eth port
-
There is also a multicast section in the settings area where you can specify your multicast interface, make sure that’s set to eth0
-
@abos_systemax Yesterday I was wrong when saying that Tom has changed some interface detection code as this was not related to the bandwidth stuff in question.
So I am just trying again: Moving this to bug reports as I remember @Tom-Elliott having changed the interface configuration/discovery code not long ago.
