Re:Deploy a Redis Cluster on Kubernetes within 5 minutes
by k_roy on Wednesday September 11, 2024 @08:38PM
(#41516561)
> Every db system has very specific nuances in data storage, io optimization, bandwidth for requests and backups, etc. a one size fits all operator project seems like it will invite trouble.
And these are all fairly well covered with standard kubernetesisms
Re:Deploy a Redis Cluster on Kubernetes within 5 minutes
by supersudo on Wednesday September 11, 2024 @11:24PM
(#41517282)
Not a redis guy, but I worked on MySQL for quite a long time. the interesting part in KubeBlocks, in my point of view, is they wrap engine-specific differences in 'Addon' . And I checked their impl for MySQL Addon.
I like their idea on hwo to modeling datebase topologies and behaviors. for instence, parameter configurations. seems they have tuned parameters for MySQL replication clusters (e.g. they render parameters w.r.t resources)
Not exactly the example I would hope for, in fact it seems duplicitous as a choice.
Databases are generally not what k8s is intended for: rapidly scalable stateless servers. There's real concerns with operators I've seen as to all the different conditions (expansions, failover, backups, detailed health checks, persistent disk management, version upgrades, boundary conditions like full disks, and if course encryption headaches).
I am not a k8s expert, I was involved with an attempt to migrate Cassandra to k8s and did not like the operator I saw at the time.
Every db system has very specific nuances in data storage, io optimization, bandwidth for requests and backups, etc. a one size fits all operator project seems like it will invite trouble.