[Design] v0.1.0 breakdown

[bjosv]

Summary

A breakdown of work items needed to fulfill requirements for a v0.1.0 release.

⚠️ Work in progress This discussion thread will be heavily modified until we find a common view.
Items will possibly be removed, replaced, broken down, and then fleshed out with details.
We will also find out if Discussions is the right tool for this.

Motivation

N/A

Detailed Design

• Genererate the initial kubebuild scaffolding. DONE

• Intial v1alpha1 valkeyCluster CRD.

• Basic support of Valkey Cluster start/stop (create/delete of CRD).
  
  The simplest reconciler to be able to create and delete a Valkey cluster (intra cluster usage)

  • Add a Service, used by a client for the inital slotmap.
  • Mount minimal Valkey config.
  • Start Valkey nodes (Deployments, non-TLS).
  • Add liveness/readiness probes.
  • Create utils to get Valkey Cluster status (convergence check).
  • Create Valkey Cluster (cluster meet).
  • Basic e2e testing.

• Support single Valkey pod kill/crash

  • Restarted pod should (still) be part of Valkey cluster
  • Add tests

• Add EventRecorder and useful events.

• Emit conditions via ValkeyClusterStatus.

• Support configuring Resources, Tolerations, NodeSelector, Affinity.

• Access a Valkey cluster from outside K8s cluster

  • Ingress/LoadBalancer/Proxy
  • cluster-announce-ip/name

• Add metrics exporter

  • ServiceMonitor, Service, create sidecars and labels.
  • CRD changes: Put ExporterResources & ExporterImage in own type.

• Add TLS

• Anti-affinity

  • A primary and its replica should be placed on different K8s worker nodes.
  • Primaries should be spread evenly.

• Support CRD updates

  • Valkey version upgrades
    • Use procedure described in https://valkey.io/topics/cluster-tutorial/#upgrade-nodes-in-a-valkey-cluster
  • Scale-in/out

Scoped out from v0.1.0

• Support Authentication
  • AUTH/ACL

Open Questions

TBA

References

TBA

[jdheyburn]

From the Friday call, the below appears to be done

• Intial v1alpha1 valkeyCluster CRD.
• Basic support of Valkey Cluster start/stop (create/delete of CRD).
• Support single Valkey pod kill/crash
  • Add tests still pending? @bjosv

These are still pending, so I have created tickets for these.

• Add EventRecorder and useful events.
  • (feat): EventRecorder implementation #27
• Emit conditions via ValkeyClusterStatus.
  • (feat): Emit conditions via ValkeyClusterStatus #26
• Add metrics exporter
  • (feat): Add Metrics Exporter Sidecar to ValkeyCluster Pods #28
  • I omitted management of Service and ServiceMonitor from this for now
• Support configuring Resources, Tolerations, NodeSelector, Affinity.
  • Resources
    • (feat): Support Resources Configuration for ValkeyCluster #29
  • Tolerations
    • (feat): Support Tolerations Configuration for ValkeyCluster #30
  • NodeSelector
    • (feat): Support NodeSelector Configuration for ValkeyCluster #31
  • Affinity
    • (feat): Support Affinity Configuration for ValkeyCluster #32
• Anti-affinity
  • A primary and its replica should be placed on different K8s worker nodes.
  • Primaries should be spread evenly.
  • ISSUE TBC

I did not create issues for these because I lack complete context, we can chat about them at the next meeting (19th Dec):

• Add TLS
• Access a Valkey cluster from outside K8s cluster
  • Ingress/LoadBalancer/Proxy
  • cluster-announce-ip/name
• Support CRD updates
  • Valkey version upgrades
    • Use procedure described in https://valkey.io/topics/cluster-tutorial/#upgrade-nodes-in-a-valkey-cluster
  • Scale-in/out

[bjosv]

Yes, the test for killing pods is still pending. I'll try to clean it up and make that PR available.