apport information
** Tags added: apport-collected uec-images xenial
** Description changed:
Hi,
I have a cloud running xenial/mitaka (with 18.02 charms).
Sometimes, an instance will take minutes to delete. I tracked down the
time taken to be file deletion :
Apr 23 07:23:00 ho
By the way the task blocks on the following :
==> /proc/54255/task/54255/stack <==
[] wakeup_preempt_entity.isra.6+0x7c/0x90
[] __switch_to+0x1f8/0x350
[] get_request+0x29c/0x910
[] blk_queue_bio+0x164/0x500
[] generic_make_request+0x154/0x310
[] submit_bio+0xd4/0x1f0
[] ext4_io_submit+0x7c/0xb0
[
Raising a task for the kernel for an opinion on how to triage this
problem further; Marking the nova task as Medium but leaving as New for
now until we figure out what's going on - this is not a show stopped but
is a pain.
** Also affects: linux (Ubuntu)
Importance: Undecided
Status: New
Thanks for continuing to dig into this.
Comparing your nova-compute logs and the strace unlink details, the
timestamps do seem to point to unlink causing the delay. Pasting the
strace unlink output below inline with the nova-compute logs in order of
occurrence:
nova-compute log:
2018-04-25 14:48:
OK, the bug happened again with strace attached to nova-compute. Once
again, there's little to no IO/network while it happens. memory is
stable. CPU is at least 50% idle (and the rest of it largely user mode).
Nothing in dmesg.
nova-compute logs are as follow :
2018-04-25 14:48:04.587 54255 INFO n
Hi Corey,
* I assume nova-compute is up the whole time because its PID doesn't change
* I'm running with debug on
* I'm going to try getting more details with strace, but the problem is I can't
repro the problem - it just sometimes happen...
No luck on repro-ing on queens so far, but the cloud i
Hi Junien,
Thanks for reporting this. It would be great to see if it's reproducible on
queens. A couple of questions:
* Is nova-compute actually up the whole time this is happening?
* Are you running with debug on?
* Could we get any more details by attaching strace to nova-compute?
Thanks,
Core
Also, nova-scheduler or nova-api-os-compute will log the following lines
(a few times per minute) while this is happening :
Apr 23 07:24:47 juju-8c74e6-4-lxd-7 nova-scheduler[15786]: 2018-04-23
07:24:47.785 15786 DEBUG nova.servicegroup.drivers.db [req-
1573c400-116c-4825-b108-3291a014b0e9 bc0ab05
