R80.20 multi queue problem - interface pending reboot

[Richard_Carson]

We a gateway and post reboot an interface that was previously working in MQ had popped out and was showing as 'pending on'

Active ixgbe interfaces:

eth1-01 [On]

eth1-02 [On]

eth1-03 [On]

eth1-04 [Pending On]

It is  under investigation by TAC as to the cause of this.

However - does any one know a way of monitor the status of interfaces in MQ so that we can alert (trap\syslog) if this scenario has occurred?

[PhoneBoy]

My question is: How did you get that information to begin with?
You could probably write up a quick monitoring script and send an snmptrap or whatever when the condition occurs.

[Richard_Carson]

cpmq get

It does look like a script is the way to go, mq status not current available with any out of the box monitoring

[Kaspars_Zibarts]

@Richard_Carson did you get a response why did it happen? I just got the same on R80.30 T215! Added more cores to MQ (10>12) and after that one of the interfaces would stay on Pending ON state

[Richard_Carson]

Hi Kaspars

It was to do with the number of IRQs available to the kernel assign being exceeded. We fixed it by removing any unused interfaces that take up a IRQs assignments even though they are not being used (i.e expansion card interfaces), after doing this problem was resolved. Not helpful if you do not have unused cards to remove!

I believe this is addressed in the new kernel but not had the opportunity to find out yet

--risc

[_Val_]

You need 3.10 kernel. And most probably Dynamic Split with it 🙂