[OmniOS-discuss] dladm hanging when creating an aggr after ramping of mtu

[Stephan Budach]

Am 09.09.15 um 16:30 schrieb Dan McDonald:
>> On Sep 9, 2015, at 10:23 AM, Stephan Budach <stephan.budach at JVM.DE> wrote:
>>
>> Is there any option to get rid of dladm without taking down the whole box?
> First, see what it's locked on:
>
> 	pstack `pgrep dladm`
>
> if you know the PID of the hung process, substitute it for the argument to pstack.
>
> I'm assuming you've tried killing this process?
>
> Dan
>

Yeah, I tried that and guess what, just right after hitting the send 
button dladm vanished. However, it had done something already resulting 
in me being unable to create aggr0 again, since it's already present, 
alas not visible.

Somehow, this seems to strike, if I try to create an aggregation on a 
card that has already one port in use, in this case ixgbe0 is already up 
and running. So, I just tried to create aggr0 only from ixgbe3, which 
went smoothly, however adding ixgbe1 to aggr0 has resulted in dladm not 
returning, so I am going for a pstack now…

…well, that was that. Unfortuanetly, right when I got the pstack, which 
basically only showed the command line I called dladm with…

dladm add-aggr -l ixgbe0 aggr0

My targets on the other end went nuts amd I had to perform some work on 
those cluster behind those… It actually ended up in booting the 014 node 
and after the reboot aggr0 had been configured. I do have some more of 
those boxes left, where I want to ramp up the mtu size and configure 
link aggregation afterwards, I am sure, I can make that happen again… ;)

Cheers,
Stephan