Pod Disruption Budget for slice gateway pods

[bharath-avesha]

🔖 Feature description

Create and reconcile PodDisruptionBudget for slice gateway pods.

🎤 Pitch

We need this enhancement to avoid connectivity loss between clusters of a slice during disruptions. Atleast one pair of slice gateway pods should be present between a pair of clusters.

✌️ Solution

We should create PodDisruptionBudget objects for the slice gateway pods, and reconcile them periodically.

🔄️ Alternative

No response

👀 Have you spent some time to check if this issue has been raised before?

• [X] I checked and didn't find similar issue

🏢 Have you read the Code of Conduct?

• [X] I have read the Code of Conduct

[Bhargav-InfraCloud]

Hi, I want to give it a try, Can you please assign this to me? /assign

[narmidm]

sure @Bhargav-InfraCloud, please let us know if you require any help.

[Bhargav-InfraCloud]

Approach from a High-level

The reconciler should watch for SliceGateway objects if:

• Created: Create a PDB with
  • matchLabels with labels that match both slice gateway's deployments. Eg. "kubeslice.io/slice=water && kubeslice.io/slicegw=water-kind-kubeslice-worker-1-kind-kubeslice-worker-2"
  • minAvailable = 1
  • labels will contain the SliceGateway's name to make it easy to delete PDB.
• Updated: There is nothing to change in PDB as the SliceGateway's name is expected to be the same even after the update.
• Deleted: Delete the matching PDB.

@narmidm @bharath-avesha If that looks reasonable, I need your opinion on the following.

1. Add the PDB logic in SliceGwReconciler that exists.
2. Create a new reconciler for creating PDB alone.

I don't think there is a great advantage to choosing one over the other. WDYT?

[narmidm]

A better option would be to use the existing SliceGwReconciler, as creating a new one could be inefficient or an overhead. You just need to ensure that it is a Requeued reconciliation and update it only when required for better efficiency. cc: @bharath-avesha