15:34:42 <Smithx10> I did dladm add-aggr and its hung
15:39:24 <Smithx10> tried killing it,  nothing,  pstack came up empty
15:44:23 <rmustacc> Do you have its process id?
15:44:45 <rmustacc> Run mdb -k, and then '::pgrep dladm | ::walk thread | ::stacks'
15:45:50 <Smithx10> https://gist.github.com/Smithx10/0002cca231fbb8374b97ae04c1f4b9f0
15:46:32 <rmustacc> Well, there's your answer to where it is.
15:46:42 <rmustacc> And why it can't be killed.
15:58:59 <danmcd> I've seen this sort of thing a lot more on i40e, but ixgbe too?
15:59:11 <Smithx10> do i just bounce it?
16:01:10 <danmcd> I might have a little-hammer solution.  Hang on...
16:01:53 <danmcd> Shell script gist coming up... alter it if you need.
16:02:28 <danmcd> https://gist.github.com/danmcd/b4a4a826be901695e0a8e2ee9682e2f0
16:02:47 <danmcd> I found this out indirectly from (I think) Jerry J.
16:03:10 <danmcd> It forces ZFS to free, which *should* free up large chunks of contiguous that devices like i40e (and apparently in your case ixgbe) need.
16:04:41 <danmcd> This should unclog ixgbe's quest for contiguous memory.  Please report back with progress?
16:23:25 <danmcd> @Smithx10 --> any luck?
17:47:44 <Smithx10> Sorry about that, was pulled into a meeting
17:49:02 <Smithx10> looks like its stuck at pwait
18:04:05 <Smithx10> looks like it took awhile but is went away :)
18:10:09 <danmcd> Gotcha.  memory reclaim is always annoying.
18:24:29 <danmcd> Smithx10: that script should be treated like a hammer (because other things will suffer when ZFS goes into a full-reclaim mindset).
23:02:57 <barfield> exit
23:02:57 <barfield> exit
23:06:59 <Smithx10> EXIT!
23:15:45 <danmcd> "in the front and in the back of the theater."