Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Related issues with pin nodes #133

Closed
RobertKrawitz opened this issue Jul 27, 2021 · 4 comments
Closed

Related issues with pin nodes #133

RobertKrawitz opened this issue Jul 27, 2021 · 4 comments

Comments

@RobertKrawitz
Copy link
Member

RobertKrawitz commented Jul 27, 2021

I'm opening this as a single issue; all of this is tightly bound up together.

1. "Pin node" is not a standard Kubernetes/OpenShift term. "Node selector" would be a better term, as that's what it is.
2. It should not be necessary to assign the benchmark operator pod to a node selector; let the scheduler take care of that.
3. It shouldn't really be necessary to assign the workload to a node selector. For cases where it matters, it should be optional.
4. For workloads that should run on different nodes, pod anti-affinity would be a better construct and could be applied by the operator itself (I suspect it already does).

@ebattat
Copy link
Collaborator

ebattat commented Jul 28, 2021

  1. You right, but this is the name convention that use in Benchmark-Operator tool in all the workloads, we need to discuss it with them.
  2. This is already done. see in the readme PIN_NODE_BENCHMARK_OPERATOR environment variable
  3. We decided that node selector will turn on by default but we can control it through configuration file but true/false parameter, i.e.
    uperf template
  4. Lets discuss together and see what is the best way to use node selector feature. one option is to pass environment variable that called selector for enable/disable this feautre.

@RobertKrawitz
Copy link
Member Author

  1. If it's inherited from upstream, I withdraw this.
  2. I don't think this (or any of the other PIN_NODE) options should be mandatory at all.
  3. I don't understand?
  4. Sure. In general, I think that all of the PIN_NODE variables should be optional. If there are situations where they shouldn't be, I'd like to analyze them.
  5. I think that these should all be command line arguments, not environment variables, though. These are all options for a specific instance of an application (benchmark-runner), not common settings that should apply across everything, such as search path, display, etc. KUBEADMIN_PASSWORD and KUBECONFIG may qualify as the latter, but IMO the others don't.

@ebattat ebattat pinned this issue Jul 28, 2021
@ebattat
Copy link
Collaborator

ebattat commented Aug 1, 2021

I have moved all the nodes to be optional.

@ebattat
Copy link
Collaborator

ebattat commented Aug 4, 2021

Please let me know if there is more issues related nodes.

@ebattat ebattat closed this as completed Aug 4, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants