Unable to boot large instances due to prlimit setting
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Cinder |
Fix Released
|
Undecided
|
Eric Harney | ||
OpenStack Compute (nova) |
Fix Released
|
High
|
Sean Dague |
Bug Description
I recently had the need to migrate some instances from an old KIlo cluster to a new Ocata one. Some of the snapshots were 120GB or more (terrible I know). Due to a prlimit limitation of cpu=8, these instances are unable to spawn.
Changing nova/virt/images.py line 42 from
cpu_time=8,
to
cpu_time=16,
allowed the instances to boot properly.
This was implemented at 2 seconds and later changed to 8 seconds as part of:
https:/
Here's my qemu-img info process taking more than 8 seconds:
9ddeea47df89414
/usr/lib/
2017-07-19 19:47:42.849 7 DEBUG oslo_concurrenc
[req-7ed3314d-
ff236d57a57dd42
9776d48734a24c2
oslo_concurrenc
LANG=C qemu-img info
/var/lib/
returned: 0 in 8.639s execute
/usr/lib/
Would it be possible to increase the default setting, or better yet make it a configuration variable so we don't have to keep chasing it?
tags: | added: low-hanging-fruit |
Changed in nova: | |
status: | New → Triaged |
importance: | Undecided → High |
I think bumping the default to something higher is a good call. This is mostly a backstop DOS prevention measure to ensure those processes end.