Description of problem:
- Create a new deployment config which will result in a pod with a dump loop.
- Attempt to scale that deployment down to 0 pods.
- The web console prevents you from using the up/down arrow widgets to accomplish this during backoffs (the predominate state during a dump loop).
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Add mysql-ephemeral template to a project.
2. Set "MySQL Database Name" to something fun, like "test; SELECT * FROM Users"
3. Instantiate the template.
4. Wait for the pod to enter a warning state.
5. Try to scale down the deployment.
I receive a no-smoking mouse cursor when hovering over the down arrow and the scale down operation is not performed.
I should be able to scale down the RC since I can do so by altering YAML.
I noticed that if I scaled down using YAML and then scaled back up using the GUI, the web console would then allow me to scale down via the GUI -- even in the warning state. This is why I suggest the issue may only affect newly created deployments.
Justin, was the deployment still in progress when you tried to scale? We disable scaling for in progress deployments. I suspect OpenShift was waiting for a readiness probe to pass before completing the deployment.
> I receive a no-smoking mouse cursor when hovering over the down arrow
In our defense, smoking _is_ bad for you.
Created attachment 1189417 [details]
Unable to scale down
Samuel, I've attached a screenshot, if it helps. The GUI overview shows the deployment is "In Progress". Clicking to view the underlying deployment# shows it to be Running.
This "In Progress" state lasts indefinitely in the GUI overview until you take an action to disrupt the deployment (e.g. like scaling down using YAML or canceling the deployment manually).
> In our defense, smoking _is_ bad for you.
I appreciate your concern ;-)
We disabled scaling in progress deployments to address this bug:
Adding Michail for input. Maybe the right solution is to let users cancel deployments from the overview.
At some point in the future we are going to enable scaling running deployments but currently it's not possible due to a backwards-compatibility issue with scaling from older clients.
Is there any use case for scaling down to zero vs using the Recreate strategy? I can't think of any.
So we have dealt with the usability problem by letting you cancel the deployment from the overview
Verified on oc/openshift 18.104.22.168
"cancel" button works on overview page for cancelling deployment.
Scale up/down arrows are valid if there is at lease 1 Active deployment When latest deploy is cancelled.
Scale up/down button will be disabled when if the only 1 running deploy cancelled.
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.