Understanding the Kubernetes Deployment Controller: Handling Resources in NewDeploymentController

Understanding the New Deployment Controller in Kubernetes

In Kubernetes, the NewDeploymentController plays a crucial role in managing the lifecycle of deployments. It acts as an event handler that responds to resource changes, specifically focusing on three main functions:

1. Handling add/update/delete operations for Deployments.
2. Managing add/update/delete operations for ReplicaSets.
3. Deleting Pods.

Question: Why is there only a delete function for Pods?

At first glance, it may seem inconsistent that the NewDeploymentController only registers a delete function for Pods while it handles add and update functions for Deployments and ReplicaSets. The absence of add/update operations for Pods suggests that these actions are likely managed elsewhere—specifically, by the ReplicaSet controller.

To dig deeper, we can refer to the ReplicaSet controller implementation. This reinforces the idea that the Deployment controller primarily oversees the ReplicaSets, which in turn manage the Pods. Thus, the Deployment controller indirectly manages Pods through the ReplicaSets.

Why is the deletePod function necessary?

The presence of the deletePod function raises another question: what purpose does it serve if the ResourceEventHandler merely checks and does not modify the resources? The key insight here is that during the process of redeploying, specifically with the RecreateDeploymentStrategyType, it is critical to ensure that all existing Pods are terminated before new ones are created.

According to Kubernetes Issue #32567 and PR #36748, the requirement states that at most one instance of user Pods should run at any given time during a deployment recreation. If a Pod is still terminating, it may not have fully stopped before the new Pod comes up. This is why the deletePod function is crucial: it ensures that the old Pod is entirely deleted before proceeding with the recreation of new Pods.

The code snippet from the NewDeploymentController illustrates this:

if d.Spec.Strategy.Type == apps.RecreateDeploymentStrategyType {
    // Sync if this Deployment now has no more Pods.
    rsList, err := util.ListReplicaSets(d, util.RsListFromClient(dc.client.AppsV1()))
    if err != nil {
        return
    }
    podMap, err := dc.getPodMapForDeployment(d, rsList)
    if err != nil {
        return
    }
    numPods := 0
    for _, podList := range podMap {
        numPods += len(podList)
    }
    if numPods == 0 {
        dc.enqueueDeployment(d)
    }
}