IPAM Helm Deployment strategy should be recreate

[braunsonm]

Description

In the helm chart for the IPAM controller, it uses the default rolling update strategy. This is a problem for two reasons:

1. The PVC should only be accessed by one pod at a time because multiple editing the same DB will cause issues
2. Not all PV types support multiple binds at the same time.

Steps To Reproduce

1. Deploy IPAM with the helm chart using a storage class that only supports one bind at a time
2. Upgrade the existing deployment
3. Notice how the deployment will never complete because the new pod cannot startup as an existing one is holding the PVC

Expected Result

Upgrades should happen transparently.

Actual Result

You need to scale the deployment to 0, then back up to 1

[trinaths]

Created [CONTCNTR-3678] for internal tracking.

[vklohiya]

Closing the issue as already fixed.

[braunsonm]

@vklohiya why is this not the default? The current setting can cause corruption and the PVC type doesn't allow multiple read/write.