f48d351c7f
Currently the script (affine-tasks.sh) waits until critical platform
pods are recovered and will then affine all platform tasks back to
the provisioned platform cpus. This leads to periods of time of 1-2
minutes during upgrade activation where platform cpus are running
close to 100% and this contributes to rabbitmq problems.
This change is to affine platform tasks only when platform upgrade
is activation-complete on a active-controller.
This removes workerconfig.service dependency as it is not present on
AIO and the start() of the script exits if the configuration is not AIO.
Test Plan:
PASS: Perform platform upgrade stx7 to stx8 in AIO-SX, AIO-DX and
confirm that all platform tasks are affined to platform cores
after the step activation-complete
PASS: Performed complete platform upgrade, sucsessfully unlocked
the node.
PASS: Install AIO-DX + worker, confirm Bootstrap completes and
standalone worker node is unlocked.
PASS: Built package successfully.
Closes-Bug: 2038537
Signed-off-by: Jim Gauld <james.gauld@windriver.com>
Change-Id: Ifc863587f8ffaf6b6478ae9edc352449430824b7
Co-authored-by: Saba Touheed Mujawar <sabatouheed.mujawar@windriver.com>
Co-authored-by: Jim Gauld <james.gauld@windriver.com>
|
||
|---|---|---|
| .. | ||
| centos | ||
| debian | ||
| opensuse | ||
| worker-utils | ||
| .gitignore | ||