New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
podpools: fix synchronizing all running containers at start #965
Conversation
Codecov Report
@@ Coverage Diff @@
## master #965 +/- ##
=======================================
Coverage 33.68% 33.68%
=======================================
Files 61 61
Lines 9156 9156
=======================================
Hits 3084 3084
Misses 5789 5789
Partials 283 283
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
pkg/cri/resource-manager/policy/builtin/podpools/podpools-policy.go
Outdated
Show resolved
Hide resolved
This change makes podpools start similar to the balloons policy start: find all running containers in the system.
fd68471
to
90b28a5
Compare
verify-log-vs-metrics pod5:pod5c1 0 20 | ||
# There should be kube-apiserver, etcd etc. running on reserved CPUs as well, | ||
# therefore allow a lot of CPU usage yet pod5 is not doing anything. | ||
verify-log-vs-metrics pod5:pod5c1 0 100 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is this somehow related?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, strongly. Now that the fix in policy Start() makes sure that kube-apiserver and etcd (among others) will be pinned to the reserved CPU - the same CPU where pod5 is scheduled too - the CPU load of that CPU can be much higher than at the time when this test was written. I got one failed test run because of that: the two processes may easily cause > 20% load on the only reserved CPU in our e2e VMs.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks 👍
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lgtm
This change makes podpools start similar to the balloons policy start: find all running containers in the system.