armadaproject / armada

A multi-cluster batch queuing system for high-throughput workloads on Kubernetes.
https://armadaproject.io
Apache License 2.0
435 stars 127 forks source link

Feature: Ideal Getting Started Experience #3677

Open csantanapr opened 3 weeks ago

csantanapr commented 3 weeks ago

I'm trying to figure out the order that will be in the new quickstart, hope this is on the website by June 20th This is my ideal workflow so far from all the different places I have read:

  1. User goes to the website https://armadaproject.io/
  2. Clicks Documentation -> dropdown selects "Getting Started" first option. This is not here today
  3. This selection takes you to https://armadaproject.io/quickstart
  4. This page have these parts
  5. Make sure you have this (kind version x+, helm, make)
  6. Run this one command (curl https/... | sh) to install armadatactl
  7. Run this commands create kind cluster
  8. Run this command to install dependencies (pulsar, prosgres, redis) using helm
  9. Run this command to install armada operator using helm
  10. Run this command to install armada instance using CRs (not CRDs, CRDs were installed with helm armada operator chart)
  11. Use Armada Now
  12. Follow this steps to configure armadactl to point to your armada instance
  13. Follow this steps to send jobs to armada
  14. Access UIs to see what ran
  15. Access Lookout UI
  16. Access Grafana UI
  17. Read here for more scenarios
dejanzele commented 2 weeks ago

Hi @csantanapr

Thank you for all the suggestions and patience with Armada, you have helped us a lot to improve our installation experience!

We have merged following PRs which should improve the installation experience:

website - https://github.com/armadaproject/armada/pull/3719 operator - https://github.com/armadaproject/armada-operator/pull/305 armada - https://github.com/armadaproject/armada/pull/3670

The new Quickstart is tested on OSX and Ubuntu.

This is the new official Quickstart - https://armadaproject.io/quickstart

csantanapr commented 2 weeks ago

I tried the new quickstart today but hit a problem creating a queue

The command to create queue

armadactl create queue example

This is the error message

Error: [armadactl.CreateQueue] error creating queue example: create queue request failed: rpc error: code = DeadlineExceeded desc = latest balancer error: connection error: desc = "error reading server preface: read tcp 127.0.0.1:65441->127.0.0.1:30002: read: connection reset by peer"
csantanapr commented 2 weeks ago

I had a problem with disk space on for my kind setup, after fixing disk space problem I tried again and queue got created

$ armadactl create queue example                                                                                                                                                                                                                                             

Created queue example

But then submitting job failed

$armadactl submit dev/quickstart/example-job.yaml                                                                                                                                                                                                                           

[Error]
Error: error submitting request &api.JobSubmitRequest{Queue:"example", JobSetId:"job-set-1", JobRequestItems:[]*api.JobSubmitRequestItem{(*api.JobSubmitRequestItem)(0x140006768c0)}}: rpc error: code = PermissionDenied desc = could not find queue "example"
Usage:
  armadactl submit ./path/to/jobs.yaml [flags]
csantanapr commented 2 weeks ago

After waiting a few minutes and retry the job then it worked

Do I need to wait for queue to be created before submitting job, how can I check the queue is ok with armadactl ?

$armadactl submit dev/quickstart/example-job.yaml                                                                                                                                                                                                                           

Submitted job with id 01j0vqkxstpdb2yvypawh2qv1w to job set job-set-1