[00:06] <lost1nfound> hey guys, i have a kernel crash and ifquery crash happening on ubuntu15.10 w/ 4.2.0-22. i need to gather coredumps for a proper bug report.. ive set up kdump per ubuntu instructions (https://goo.gl/3vXG4g), and i cant get kdump to write a vmcore no matter what i try. ive tried adjusting crashkernel=memory to everything between 128M and 1G, ive tried configing kdump to dump over ssh but i never s
[00:06] <lost1nfound> ee it connect and no dump, very stuck :)
[00:07] <lost1nfound> any ideas what to check/where to look to get my vmcore dumps working?
[00:08] <lost1nfound> also this is on EC2, so i have no way to get a full read/write console. so i cant just make kdump/kexec drop me into a shell and do it manually or anything
[00:08] <lost1nfound> crash seems to be happening inside the network driver also which seems to be a further debugging complication
[00:09] <lost1nfound> wanting to submit a useful bug report but i need a vmcore for that
[00:39] <lost1nfound> and so ive read the reports that there are problems if /boot is a different partition, but my entire dirstructure is on my root volume, no other disks or partitions attached. only potential complication might be that this is an EBS disk on ec2? but even then id expect kdump over ssh to work and it doesnt
=== shuduo-afk is now known as shuduo
[04:40] <hallyn> sforshee: bug 1530617 - this is a funky one.  i suspect i'll need to use taskstats or systemtap to track it down.  if you have any better ideas for how to debug pls shout
[04:40] <ubot5`> bug 1530617 in lxc (Ubuntu) "FUSE in wily image with upstart installed causes chaos" [High,Confirmed] https://launchpad.net/bugs/1530617
[05:30] <lost1nfound> anyone know anything about getting kdump to work when its not dumping?
[07:47] <hallyn> meh, systemtap is failing to build modules both in wily and xenial
[13:23] <sforshee> hallyn: no better ideas off the top of my head, I played with it a while yesterday but didn't learn much. If I think of something I'll let you know.
[13:25] <sforshee> hallyn: huh, the same thing happened just now when I ran losetup in the host. Maybe it has something to do with uevents?
[13:26] <sforshee> hallyn: and again when I mounted that lop device in the host
[13:43] <rtg> sforshee, is bug #1528688 on your to-do list ?
[13:43] <ubot5`> bug 1528688 in linux-firmware (Ubuntu) "scsi-firmware udeb does not include ql2500_fw.bin" [Undecided,In progress] https://launchpad.net/bugs/1528688
[13:46] <sforshee> rtg: I've been watching it, the last comment said to hold off on the patches but looks like he just addded a new one
[13:46] <sforshee> so I'll get that done today
[13:47] <rtg> ah, I guess I missed that
[14:16] <rtg> apw, been looking at bug #1530540 - do you remember if we had a good reason for why we put the thermald Depends/Recommends in the meta package instead of the actual package ?
[14:16] <ubot5`> bug 1530540 in linux (Ubuntu) "Wish: Put thermald as recommendation into linux-image-*.*.*-*-generic, not into linux-image-generic" [Wishlist,Triaged] https://launchpad.net/bugs/1530540
[14:20] <apw> rtg, because we use those packages for linux-virtual for which thermald and crda etc are not sensible
[14:21] <rtg> apw, makes sense. I knew there was a reason we did it that way.
[14:22] <apw> the question is why they care to want it there
[14:22] <rtg> apw, nit picking ?
[14:23] <rtg> anyways, I'll respond
=== ubot5` is now known as ubot5
[14:55] <sforshee> hallyn: I'm definitely seeing a direct correlation between bug 1530617 and uevents, any command I run in the host that causes uevents to be generatted also causes processes in the container to be killed
[14:55] <ubot5> bug 1530617 in lxc (Ubuntu) "FUSE in wily image with upstart installed causes chaos" [High,Confirmed] https://launchpad.net/bugs/1530617
[16:39] <apw> rtg, thanks it doesn't seem right to change it for sure
[17:04] <hallyn> sforshee: oh, well systemtap (when rebuilt from upstream) did give me a bunch of udevs being killed,
[17:04] <hallyn> maybe udev gets killed and then upstart is upset so it goes on a rampage
[17:04] <hallyn> maybe a non-ubuntu-cloud container doesn't have the problem bc it doesnt 'have udev
[17:06] <hallyn> well that's not it
[17:21] <sforshee> hallyn: systemd has udev too though and it doesn't happen with systemd
[17:22] <sforshee> is there some way that lxcfs and upstart might interact in response to uevents that causes it?
[17:37] <hallyn> sforshee: yes, my guess is it has to do with the way upstart ptraces it's daemons
[17:37] <hallyn> but that's just a wild guess based on the fact that it only happens with upstart
[17:37] <hallyn> might be possible to write a test program to run in a userns to verify
[17:38] <sforshee> hallyn: what's the relationship with lxcfs then? Didn't you say that stopping lxcfs prevents it?
[17:39] <hallyn> oh i don't think it *prevents* the problem itself.
[17:39] <hallyn> it's just that with lxcfs running 'lxc-start -n x1; lx-attach -n x1' always gets killed at lxc-attach after a sec or 2
[17:40] <hallyn> presumably d uring the lxcfs bind mounts which dreate uevents as you found
[17:41] <sforshee> got it, I misunderstood your comment on the bug
=== henrix_ is now known as henrix
[20:57] <dobey> hi. anyone have ideas on how to figure out why the lowlatency kernel hangs, but only when intel video is driving the monitor at 30Hz, and not at 60Hz?