-
Notifications
You must be signed in to change notification settings - Fork 113
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
config-daemon: Restart all instances of device-plugin #783
config-daemon: Restart all instances of device-plugin #783
Conversation
Thanks for your PR,
To skip the vendors CIs, Maintainers can use one of:
|
f5f470b
to
f2fdee5
Compare
Pull Request Test Coverage Report for Build 11255003478Details
💛 - Coveralls |
f2fdee5
to
c318a89
Compare
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
Failing test
is addressed in |
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
When the operator changes the device-plugin Spec (e.g. .Spec.NodeSelector), it may happen that there are two device plugin pods for a given node, one that is terminating, the other that is initializing. If the config-daemon executes `restartDevicePluginPod()` at the same time, it may kill the terminating pod, while the initializing one will run with the old dp configuration. This may cause one or more resources to not being advertised, until a manual device plugin restart occurs. Make the config-daemon restart all the device-plugin instances it founds for its own node. Signed-off-by: Andrea Panattoni <[email protected]>
c318a89
to
f286a04
Compare
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!
When the operator changes the device-plugin Spec (e.g. .Spec.NodeSelector), it may happen that there are two device plugin pods for a given node, one that is terminating, the other that is initializing.
If the config-daemon executes
restartDevicePluginPod()
at the same time, it may kill the terminating pod, while the initializing one will run with the old dp configuration. This may cause one or more resources to not being advertised, until a manual device plugin restart occurs.Make the config-daemon restart all the device-plugin instances it founds for its own node.