Support parallel QEMU kola tests based on memory amount #3118
Labels
enhancement
New feature or request
Comments
|
Most tests require only 1G of RAM and only recently did some tests get bumped to 1.5G due to https://bugzilla.redhat.com/show_bug.cgi?id=1907030 (hopefully temporarily). So I don't think we should spend a lot of time on the scheduling algorithm here. Even a basic one like "wait until enough memory has been freed to schedule the next test" should suffice. |
jlebon
added a commit
to jlebon/fedora-coreos-pipeline
that referenced
this issue
Oct 12, 2022
Some of the tests require 1.5G of RAM now due to https://bugzilla.redhat.com/show_bug.cgi?id=1907030. We need to bump our calculation of parallelism accordingly in the absence of coreos/coreos-assembler#3118. Otherwise depending on how kola schedules tests, we might go over the limit.
dustymabe
pushed a commit
to coreos/fedora-coreos-pipeline
that referenced
this issue
Oct 12, 2022
Some of the tests require 1.5G of RAM now due to https://bugzilla.redhat.com/show_bug.cgi?id=1907030. We need to bump our calculation of parallelism accordingly in the absence of coreos/coreos-assembler#3118. Otherwise depending on how kola schedules tests, we might go over the limit.
jlebon
added a commit
to jlebon/coreos-assembler
that referenced
this issue
Oct 12, 2022
This roughly matches what we do now in e.g. the FCOS pipeline and a few other places. Otherwise, we may hit against our memory limit and get evicted. See coreos#3118 for more background info.
jlebon
added a commit
to jlebon/coreos-assembler
that referenced
this issue
Oct 12, 2022
This roughly matches what we do now in e.g. the FCOS pipeline and a few other places. Otherwise, we may hit against our memory limit and get evicted. See coreos#3118 for more background info.
jlebon
added a commit
that referenced
this issue
Oct 12, 2022
This roughly matches what we do now in e.g. the FCOS pipeline and a few other places. Otherwise, we may hit against our memory limit and get evicted. See #3118 for more background info.
dustymabe
pushed a commit
to dustymabe/coreos-assembler
that referenced
this issue
Nov 3, 2022
This roughly matches what we do now in e.g. the FCOS pipeline and a few other places. Otherwise, we may hit against our memory limit and get evicted. See coreos#3118 for more background info. (cherry picked from commit 5d12d39)
dustymabe
pushed a commit
to dustymabe/coreos-assembler
that referenced
this issue
Nov 3, 2022
This roughly matches what we do now in e.g. the FCOS pipeline and a few other places. Otherwise, we may hit against our memory limit and get evicted. See coreos#3118 for more background info. (cherry picked from commit 5d12d39)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Feature Request
Desired Feature
In addition to per-process parallelism, kola should support not going over a certain amount of total memory. This ensures that it meshes well in Kubernetes pods where memory limits could be enforced.
One strawman is to add a new
--qemu-max-memory MEM_IN_MIB. With this,kolawill treat--parallelas a maximum while respecting--qemu-max-memory.Example Usage
Other Information
Split out of #1438.
If there's a way for kola to auto-discover how much memory is available to the cgroup it's running in, we could have a
--qemu-max-memory autolike we do for--parallel.The text was updated successfully, but these errors were encountered: