isolcpu interactions with non-guaranteed pods

Bug #1830634 reported by Brent Rowsell
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
StarlingX
Fix Released
High
David Sullivan

Bug Description

Brief Description
-----------------
isolcpu's does not play nice with besteffort or burstable pods.

With isolcpu's enabled these pods will get stuck on the core where they were first scheduled vs floating. Overlaying exlusive cpu's with this, one can end up in a situation where the besteffort/burstable pods are running on the same cpu as the exclusive pods.

There will always be some besteffort or burstable pods on a worker node as the system pods (calico et) fall into the category.

So for worker nodes that do not have the openstack-compute label assigned we need to remove the isolcpu's boot arg. I believe the change would be at line 581 in platform.py plugin.

Severity
--------
Major

Steps to Reproduce
------------------
Launch pods with exclusive CPU's

Expected Behavior
------------------
Pods using defaultCpuSet should be able to float across that set

Actual Behavior
----------------
See description

Reproducibility
---------------
100%

System Configuration
--------------------
Low-latency worker

Branch/Pull Time/Commit
-----------------------
 BUILD_DATE="2019-05-24 17:42:34 -0400"

Last Pass
---------
Never

Timestamp/Logs
--------------
N/A

Test Activity
-------------
Other

tags: added: stx.2.0 stx.config
Revision history for this message
Ghada Khalil (gkhalil) wrote :

Marking as release gating; high priority as this can result in stability and performance issues for the containerized env

Changed in starlingx:
importance: Undecided → High
status: New → Triaged
tags: added: stx.containers
Changed in starlingx:
assignee: nobody → David Sullivan (dsullivanwr)
description: updated
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix proposed to config (master)

Fix proposed to branch: master
Review: https://review.opendev.org/661817

Changed in starlingx:
status: Triaged → In Progress
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix merged to config (master)

Reviewed: https://review.opendev.org/661817
Committed: https://git.openstack.org/cgit/starlingx/config/commit/?id=c82d9b0ec6aca7df1c72bd5c6692d4d731b3e5f5
Submitter: Zuul
Branch: master

commit c82d9b0ec6aca7df1c72bd5c6692d4d731b3e5f5
Author: David Sullivan <email address hidden>
Date: Tue May 28 13:13:27 2019 -0400

    isolcpu interactions with non-guaranteed pods

    For low latency worker nodes without the openstack compute label set,
    the isolcpu parameter should be restricted to vswitch cores (if any).

    Change-Id: I16ab59a6110fc5d8a1acfe5b6740868de3875585
    Closes-Bug: 1830634
    Signed-off-by: David Sullivan <email address hidden>

Changed in starlingx:
status: In Progress → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.