search

openebs / mayastor

Dynamically provision Stateful Persistent Replicated Cluster-wide Fabric Volumes & Filesystems for Kubernetes that is provisioned from an optimized NVME SPDK backend data storage stack.
Apache License 2.0
756 stars 109 forks source link

[Doc Source]: Quickstart is missing information for cluster preparation #1605

Closed cmontemuino closed 6 months ago

cmontemuino commented 8 months ago

Are you reporting an issue with existing content?

Section "Label Mayastor Node Candidates" from "Quickstart -> Preparing the cluster" is not complete. The requisite is to label nodes with openebs.io/engine=mayastor, but if you set csi.​node.​topology.​nodeSelector, then label openebs.io/csi-node=mayastor is also required.

The situation is a bit more complicated with the agent-ha-node. Only by inspecting the chart I could realize it is using the selectors defined under the csi.node.topology section. This is very confusing.

Are you proposing new content, or a change to the existing documentation layout or structure?

cmontemuino commented 8 months ago

I can open a PR with proposed changes if required.

tiagolobocastro commented 8 months ago

but it is also required to label them with openebs.io/csi-node=mayastor.

hmm it's not necessary, only if you want to control on which nodes to run - by default they will run on any node.

cmontemuino commented 8 months ago

but it is also required to label them with openebs.io/csi-node=mayastor.

hmm it's not necessary, only if you want to control on which nodes to run - by default they will run on any node.

Yes, you're right. I've updated the description with this note:

... but if you set csi.​node.​topology.​nodeSelector, then label openebs.io/csi-node=mayastor is also required

It's kind of OKish if it was only the CSI Controller. Having the agent-ha-node daemonset using the same labels is a bit baffling. I still believe the documentation can be improved; otherwise there's no other action possible for users than looking at the chart implementation (as I had to do).

tiagolobocastro commented 8 months ago

The agent-ha-node should run on the same nodes as the csi-nodes, but I agree it does make it confusing.

I suggest as a first step let's improve the docs; then maybe we can think about how to improve label usage somehow or even potentially joining ha-node in the same pod as the csi, though I'm not quite sure if it's a good idea to do that.

cmontemuino commented 8 months ago

@tiagolobocastro -- I've opened two PRs to address this issue:

tiagolobocastro commented 6 months ago

Completed by @cmontemuino