I'd like all of the in-flight workload APIs (ReplicaSet, Deployment, StatefulSet, DaemonSet) to advance to v1 together as a group, sometime this year.

Related issues:

• Deployment v1 issue: Deployment graduation requirements to v1/GA #20130
• DaemonSet v1 issue: DaemonSet: requirements for graduation to beta and then to v1 #15310
• StatefulSet beta: Pet Set in beta #28718
• Eliminate extensions API group: stop serving extensions/v1beta1 and networking.k8s.io/v1beta1 in 1.22 #43214
• Workload controller design issue umbrella issue: Write proposal for controller pod management: adoption, orphaning, ownership, etc. (aka controllers v2) #14961
• Extension patterns: Standardize extension patterns #31571, Implement External strategy for workload controllers #39337
• Common patterns: Common patterns in workload controllers #37563, Status field means different things for different controllers #42410
• https://github.com/kubernetes/community/wiki/Roadmap:-Deployment
• https://github.com/kubernetes/community/wiki/Roadmap:-ReplicaSet

I'd like to see at least:

• Server-side cascading deletion (GC) by default (e.g., Deployment in an inconsistent after kubectl delete and ctrl + C #23252, Deleting the deployment from UI doesn't delete the replica sets and pods #40014)
• Controller ref (Implement controllerRef #24946) to provide mutual exclusion even across controller types, but also ensure that orphaning and adoption work as expected
• Upgrades/rollouts for DaemonSet and StatefulSet (Pet set upgrades #28706)
• Decide whether to change label/selector defaulting and/or selector update behavior (Server side defaulting of values prevents diff patches from working #34292, kubectl apply deployment -f doesn't accept label/selector changes #26202, kubectl replace deployment doesn't remove ReplicaSet from the old deployment #24888, Allow Users to change labels using deployment #14894, ...)
• Decide whether to move forward with templateGeneration (Refine the Deployment proposal and move away from hashing community#384) or with the new hash (Refine the Deployment proposal and switch hashing algorithm community#477, Error syncing deployment, replica set already exists #29735)
• Decide whether to rename kubectl undo (Deployment Undo doesn't undo number of replicas #25236)
• Decide whether we're going to try to unify storage for ReplicationController and ReplicaSet
• DaemonSet and StatefulSet history and rollback
• Scale subresource for all controllers for which it makes sense
• Deployment overlap annotation cleanup (Clean up Deployment overlap annotation code #43322)
• Decide whether/which annotations should be propagated by updates (Allow kubectl annotate to annotate nested template objects #37666)
• Decide what to do about clock skew (e.g., move minReadySeconds to pod MinReadySeconds / AvailableReplicas for ReplicaSets #32771 (comment); Master/node clock skew will affect deployment's pod availability check  #29229)
• Documentation completed (e.g., Advocate Deployment instead of ReplicationController in documents once it's beta #18465)
• Double-check validation and defaults (e.g., Handle maxUnavailable larger than spec.replicas #42479)
• Fix/cleanup taint/toleration support in DaemonSet (taints & tolerations with daemonsets #29738), affinity/anti-affinity (DaemonSet should respect Pod Affinity and Pod AntiAffinity #29276), schedulability (Running DaemonSet only on Schedulable Nodes #26810), scheduling on master nodes (Daemonset is trying to schedule on master #29108), durability (Pod created by DamonSet is always in running phase  #29193), etc.
  • Maybe via DaemonSet use of scheduler (DaemonSet pods should be scheduled by default scheduler, not DaemonSet controller #42002)

Bugs:

• Scaled-down Deployments don't identify old ReplicaSets (Scaled down deployments cannot identify old replica sets #42570)

Maybe:

• Finish UnavailableReplicas (Solidify UnavailableReplicas field for Deployments #41789)
• Disallow activeDeadlineSeconds (Disallow setting activeDeadlineSeconds for restartAlways pods/controllers #38684)
• Rollout strategy update improvement (changing update strategy and "applying" is not allowed #24198, Server side defaulting of values prevents diff patches from working #34292, Add discriminators to all unions/oneof in the API #35345)
• ConfigMap rollouts (Facilitate ConfigMap rollouts / management #22368, docs: add proposal for ConfigMap management #31701, Atomic Configmaps/Secrets #43112, ...)
• Extensible rollout strategies (RFC: Support user defined and extensible deployment strategies #14510)
• Make controllers aware of namespace deletion (Controllers should be aware of namespace termination? #38612, Deleting a namespace sometimes leaves orphaned deployment and pods which can't be deleted #36891)
• Disallow label changes in status updates (Status updates should blacklist label updates #20978)
• Derive DaemonSet pod names from nodes (Services backed by DaemonSet pods should have a hostname based network identifier #41977)
• Coordinate DaemonSet updates with node updates (Proposal: versioned DaemonSet #42917) -- may need to change semantics of node selector, tolerations, and other scheduling constraints

Desirable, but can probably be done later:

• Implement observedGeneration (Create per-object sequence number and report last value seen in status of each object #7328)
• Quota support (Quota for DaemonSets and Deployments #23360)
• Ensure the controllers generate the desired events, such as events when they can't create pods (All controllers should raise events when they can't create objects as expected #21912)
• Graceful degradation (e.g., Replication controller (and RS): Don't make more than 1k unschedulable pods #33041, ReplicationController + crashlooping or invalid docker image = bad times #2529)
• Eliminate special-case hacks for the controllers (e.g., Scheduler, StatefulSets, and External Controllers #39687)
• Eliminate Expectations (Revisit use of expectations in controllers #22061)
• Performance/scalability improvements (e.g., CronJob controller should use shared informers #17130, The addNode and UpdateNode methods in pkg/controller/daemon/controller.go should only enqueue related DaemonSet #17441, WIP: Add pod-to-rc cache for RC/Job/DaemonSet/Endpoint #18972, We should use targeted watch a poll on the cache in deployment controller #34598, Index pods by node name for the daemon set controller #38868, Kubernetes API extremely slow to trigger scale/scheduling #40271)
• Automatic rollback (Automatic rollback for failed deployments #23211)
• Infant mortality detection (Rolling updates should fail early if a pod fails to start #18568)
• Auto-pause (Promotable canary deployments (aka auto-pause) #11505)
• Attach, exec, portforward, proxy (e.g., Exec should use AttachablePodForObject so that we can exec into a deployment #21227)
• Status cleanup (Status field means different things for different controllers #42410)

Definitely should not block v1/GA:

• Factor out common functionality (Common patterns in workload controllers #37563)
• Support relabeling (Provide a way to relabel controllers and their pods #36897)
• Scale down gradually (Scale down gracefully #23885)
• PodDisruptionBudget completion (DisruptionBudget object to define the max disruption that can be caused to pods #12611)
• Critical pod hack cleanups in favor of general priority/preemption (Preemption priority / scheme #22212, Guaranteed admissions and evictions of "critical pods" and "static pods" #40573)
• Easier image/tag updates (Feature request: Option to update just the image tag in a deployment #29097) / image triggers
• Hooks (RFC: Deployment hooks #14512)
• Image tag resolution (Image name/tag resolution preprocessing pass #1697)
• Bulk quota (Warn or error when a workload object (eg Deployment) would breach a quota #25146)
• In-place updates (In-place rolling updates #9043)
• Rolling restart (Rolling restart of pods #13488)
• Splitting out the pod template (Separate the pod template from replicationController #170)
• Selector overlap detection (kubectl rollout for overlapping deployments #43321, Constraint solver to determine overlapping label selectors #19830, Validate no replicationController overlap. #2210)
• Failure-domain awareness (e.g., Scheduler input should be taken when reducing replicas #4301, Deployment should support both Failure Domains and Update Domains #41394)
• Federation support (Handle updates, rolling update preferences and revisions in federated deployment controller #37570)
• Service generation (RFC: Coupled Services and Deployments #14956)
• Blue-green deployments (Support native blue/green deployments #33462)
• Min/max replica counts (Feature request: ability to specify a min/max replica count on scaleable resources #33843)
• Cron/batch jobs that run on every node (DaemonSet should allow for a RestartPolicy of OnFailure #17182, CronJob daemonset (previously ScheduledJob) #36601)

TBD:

• Better history support (Better way to communicate deployment's RS garbage collection #28373)
• Kill specific pods ([Feature]able to stop/restart/scale specific pod labled by pod-template-hash in one deployment #36868)
• Deleted controllers in limbo (Controllers don't take any actions except updating Status if deletionTimestamp != nil #24934)

We also should triage related issues that have been filed, though I did a quick triage through deployment and daemonset issues.

cc @erictune @janetkuo @kow3ns @Kargakis @foxish @enisoc @smarterclayton