-
Notifications
You must be signed in to change notification settings - Fork 18
Open
Description
When running this step in parallel workflows, a race condition on the config.lock file generated when changing the kube context is possible. This is very intermittent, but when it happens, one (or more) steps will error with: error: open /codefresh/volume/sensitive/.kube/config.lock: file exists.
This behavior can be reproduced by running two kubectl config use-context commands in parallel (e.g. two different terminals).
These calls are originating from the use-context command that is present in all install and promotion scripts.
Would it be possible to make this an optional param/create bypass param in order to support use cases where multiple releases are being installed into the same cluster, and the context has been preset?
Thanks!
Metadata
Metadata
Assignees
Labels
No labels