OOM errors in ZFS integration testing
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
curtin |
Triaged
|
Medium
|
Unassigned |
Bug Description
In two separate curtin integration test runs, we've seen ZFS tests OOM:
ERROR: test suite for <class 'vmtests.
-------
Traceback (most recent call last):
File "/usr/lib/
self.setUp()
File "/usr/lib/
self.
File "/usr/lib/
super(
File "/usr/lib/
try_
File "/usr/lib/
return func()
File "/var/lib/
'\n'
Exception: CosmicTestZfsRo
[ 262.860664] Out of memory: Kill process 938 (cloud-init) score 29 or sacrifice child
[ 262.861969] Killed process 973 (python3) total-vm:41748kB, anon-rss:14132kB, file-rss:0kB, shmem-rss:0kB
and
ERROR: test suite for <class 'vmtests.
-------
Traceback (most recent call last):
File "/usr/lib/
self.setUp()
File "/usr/lib/
self.
File "/usr/lib/
super(
File "/usr/lib/
try_
File "/usr/lib/
return func()
File "/var/lib/
'\n'
Exception: EoanTestZfsRoot
[ 352.997122] cloud-init[756]: Setting up linux-modules-
[ 355.370729] Out of memory: Kill process 27158 (depmod) score 24 or sacrifice child
[ 355.372005] Killed process 27158 (depmod) total-vm:54192kB, anon-rss:48584kB, file-rss:1676kB, shmem-rss:0kB
(The source for these errors: https:/
Changed in curtin: | |
status: | New → Triaged |
importance: | Undecided → Medium |
We already bump disco and eoan to 2G VMs, but if there are a lot of *updates* since the daily image was released, we may see these.
We could bump Disco/Eoan and up to 4G though I don't want to do that until we land something to address memory pressure on the host. Alternatively, we can bump to 4G in a branch and also see about reducing the total number of tests we run.