Create valkey-k8s-operator#28
Conversation
This RFC proposes an open-source Kubernetes Operator for Valkey. It supports standalone and Sentinel-based HA deployments, with optional TLS, persistence, and Prometheus metrics integration. Includes: CRD schema and operator behavior ConfigMap and Secret handling Sentinel-managed shard support AZ-aware scheduling design Optional Failover CRD Learnings from Helm-based deployments and internal prototypes Signed-off-by: Andrey G <98027999+andrey-glazkov@users.noreply.github.com>
|
I would leave Sentinel as nice to have and Cluster Mode as a must have for version 0.1.0 besides Prometheus integration with Coroot will be ideal for Observability. Failover and node replacement has to be part of the functionality as well for reliability. |
I think we don't need to define exactly what's must-have for a particular version. We can line out the long term goals, implement things incrementally and just make sure we don't shut the door on any of these long term goals. It's great to have something to start with. We should review this RFC and make it covers what people want. |
- Updated module reference to libvalkey_bloom.so - Clarified Cluster Mode as primary HA/scalable deployment - Updated Design Considerations to remove Bitnami Helm chart dependency - Retained Sentinel HA and standalone mode for smaller deployments Signed-off-by: Andrey G <98027999+andrey-glazkov@users.noreply.github.com>
Made sure we reference Valkey in all examples Signed-off-by: Andrey G <98027999+andrey-glazkov@users.noreply.github.com>
zuiderkwast
left a comment
zuiderkwast
left a comment
There was a problem hiding this comment.
Added some comments and some of our requirements as suggestions in the comments.
andrey-glazkov
left a comment
andrey-glazkov
left a comment
There was a problem hiding this comment.
adding some simple changes here re Cluster mode deployments and Requirements section
|
Would it make sense to fork Hyperspike's operator (https://github.com/hyperspike/valkey-operator) into this org and then continue development from there? It seems that the basics (cluster mode) are already supported |
Signed-off-by: Andrey G <98027999+andrey-glazkov@users.noreply.github.com>
Update Valkey Operator RFC with clarified mode definitions (cluster/sentinel/standalone), reliability requirements, TLS/mTLS support, Prometheus exporter wording, and CRD strategy. Added requirements section and design overviews for all modes. Signed-off-by: Andrey G <98027999+andrey-glazkov@users.noreply.github.com>
added .md Signed-off-by: Andrey G <98027999+andrey-glazkov@users.noreply.github.com>
| @@ -0,0 +1,219 @@ | |||
| RFC: 21 | |||
There was a problem hiding this comment.
The RFC number is the PR number, which is #28
| RFC: 21 | |
| RFC: 28 |
8 participants
This RFC proposes an open-source Kubernetes Operator for Valkey. It supports Cluster and Sentinel-based HA deployments and standalone nodes, with optional TLS, persistence, and Prometheus metrics integration.
Includes:
CRD schema and operator behavior
ConfigMap and Secret handling
Sentinel-managed shard support
AZ-aware scheduling design
Optional Failover CRD
Learnings from Helm-based deployments and internal prototypes