Minimum memory requirements for the example cluster to start

[diegonayalazo]

Hi Knative community, thanks for the quickstart plugin. It saves a lot of time setting up the cluster. It would be great to know the amount of memory it will consume beforehand. My laptop is resource limited (8GB, not expandable) and having an IDE, plus the cluster crashed my machine in Linux or sometimes it takes the plugin (kn quickstart kind) up to 20 minutes to start up.

Thank you for all your support!

[psschwei]

Related docs issue: https://github.com/knative/docs/issues/4245

[github-actions[bot]]

This issue is stale because it has been open for 90 days with no activity. It will automatically close after 30 more days of inactivity. Reopen the issue with /reopen. Mark the issue as fresh by adding the comment /remove-lifecycle stale.

[psschwei]

/remove-lifecycle stale

[csantanapr]

Right now because of the requests configured in the quickstart set of resources plus the tutorial resources, the total CPU request is ~2.3 and memory is ~1.4Gi

The actual utilization of the cluster is way below about ~0.3 CPU and ~1 Gi memory

Here is the output of cluster I created using quickstart

❯ kubectl resource-capacity --sort cpu.request --pods  --util
NODE                    NAMESPACE            POD                                                    CPU REQUESTS   CPU LIMITS     CPU UTIL    MEMORY REQUESTS   MEMORY LIMITS   MEMORY UTIL

knative-control-plane   *                    *                                                      2305m (76%)    4600m (153%)   258m (8%)   1440Mi (72%)      4590Mi (230%)   4Mi (0%)
knative-control-plane   knative-serving      activator-85bd4ddcbb-54kdl                             300m (10%)     1000m (33%)    2m (0%)     60Mi (3%)         600Mi (30%)     16Mi (0%)
knative-control-plane   kube-system          kube-apiserver-knative-control-plane                   250m (8%)      0Mi (0%)       77m (2%)    0Mi (0%)          0Mi (0%)        400Mi (20%)
knative-control-plane   kube-system          kube-controller-manager-knative-control-plane          200m (6%)      0Mi (0%)       23m (0%)    0Mi (0%)          0Mi (0%)        64Mi (3%)
knative-control-plane   knative-eventing     eventing-webhook-5968f79978-fv66l                      100m (3%)      200m (6%)      3m (0%)     50Mi (2%)         200Mi (10%)     23Mi (1%)
knative-control-plane   knative-eventing     mt-broker-ingress-5ddd6f8b5d-drsb8                     100m (3%)      0Mi (0%)       1m (0%)     100Mi (5%)        0Mi (0%)        9Mi (0%)
knative-control-plane   knative-serving      autoscaler-84fcdc5449-g2j2t                            100m (3%)      1000m (33%)    6m (0%)     100Mi (5%)        1000Mi (50%)    24Mi (1%)
knative-control-plane   knative-serving      controller-6fd5bb86df-7cjl5                            100m (3%)      1000m (33%)    4m (0%)     100Mi (5%)        1000Mi (50%)    34Mi (1%)
knative-control-plane   kube-system          coredns-78fcd69978-f5gdg                               100m (3%)      0Mi (0%)       3m (0%)     70Mi (3%)         170Mi (8%)      15Mi (0%)
knative-control-plane   kube-system          etcd-knative-control-plane                             100m (3%)      0Mi (0%)       27m (0%)    100Mi (5%)        0Mi (0%)        62Mi (3%)
knative-control-plane   knative-serving      domainmapping-webhook-8484d5fd8b-tshr8                 100m (3%)      500m (16%)     3m (0%)     100Mi (5%)        500Mi (25%)     16Mi (0%)
knative-control-plane   kube-system          metrics-server-5794ccf74d-cd5qj                        100m (3%)      0Mi (0%)       4m (0%)     200Mi (10%)       0Mi (0%)        20Mi (0%)
knative-control-plane   kube-system          kube-scheduler-knative-control-plane                   100m (3%)      0Mi (0%)       3m (0%)     0Mi (0%)          0Mi (0%)        22Mi (1%)
knative-control-plane   knative-serving      webhook-97c648588-8hgrt                                100m (3%)      500m (16%)     4m (0%)     100Mi (5%)        500Mi (25%)     19Mi (0%)
knative-control-plane   knative-eventing     mt-broker-filter-574dc4457f-r9t4q                      100m (3%)      0Mi (0%)       1m (0%)     100Mi (5%)        0Mi (0%)        11Mi (0%)
knative-control-plane   knative-eventing     mt-broker-controller-8d979648f-g4fxs                   100m (3%)      0Mi (0%)       2m (0%)     100Mi (5%)        0Mi (0%)        17Mi (0%)
knative-control-plane   knative-eventing     eventing-controller-58875c5478-x9zlm                   100m (3%)      0Mi (0%)       6m (0%)     100Mi (5%)        0Mi (0%)        29Mi (1%)
knative-control-plane   kube-system          kindnet-p55bv                                          100m (3%)      100m (3%)      1m (0%)     50Mi (2%)         50Mi (2%)       10Mi (0%)
knative-control-plane   kube-system          coredns-78fcd69978-qk7bg                               100m (3%)      0Mi (0%)       2m (0%)     70Mi (3%)         170Mi (8%)      13Mi (0%)
knative-control-plane   knative-serving      domain-mapping-74d5d688bd-tb475                        30m (1%)       300m (10%)     1m (0%)     40Mi (2%)         400Mi (20%)     13Mi (0%)
knative-control-plane   default              cloudevents-player-00001-deployment-65c477d844-4cbd5   25m (0%)       0Mi (0%)       5m (0%)     0Mi (0%)          0Mi (0%)        21Mi (1%)
knative-control-plane   local-path-storage   local-path-provisioner-85494db59d-gb44d                0Mi (0%)       0Mi (0%)       2m (0%)     0Mi (0%)          0Mi (0%)        8Mi (0%)
knative-control-plane   kube-system          kube-proxy-7qfn9                                       0Mi (0%)       0Mi (0%)       4m (0%)     0Mi (0%)          0Mi (0%)        17Mi (0%)
knative-control-plane   knative-serving      net-kourier-controller-66bc9d6697-wxwxz                0Mi (0%)       0Mi (0%)       8m (0%)     0Mi (0%)          0Mi (0%)        44Mi (2%)
knative-control-plane   knative-eventing     imc-controller-86cd7b7857-whg8v                        0Mi (0%)       0Mi (0%)       5m (0%)     0Mi (0%)          0Mi (0%)        22Mi (1%)
knative-control-plane   knative-eventing     imc-dispatcher-7fcb4b5d8c-z7rzl                        0Mi (0%)       0Mi (0%)       2m (0%)     0Mi (0%)          0Mi (0%)        15Mi (0%)
knative-control-plane   kourier-system       3scale-kourier-gateway-58856c6cc7-spj48                0Mi (0%)       0Mi (0%)       2m (0%)     0Mi (0%)          0Mi (0%)        21Mi (1%)

I was using colima as docker desktop today and the default VM uses 2CPU and 2Gi and kn quickstart didn't work because the CPU requests are higher 2.3 and some pods got stuck in pending.

@psschwei I'm thinking that we should delete the resources from all pods, this will remove all requests for CPU and memory and allow knative to run on cluster as small as 1CPU and 1Gi of memory

if you are using colima you can delete and recreate with 3CPUs

colima delete
colima start --cpu 3

Here is an example used in this github action to delete all resources

curl -L -s $base/serving-core.yaml | yq 'del(.spec.template.spec.containers[]?.resources)' -y | yq 'del(.metadata.annotations."knative.dev/example-checksum")' -y | kubectl apply -f -

In our case, I think we would need to parse the yaml and manipulate before we do the apply to avoid asking the user to install yq in their computer.

[psschwei]

I'd be a little hesitant to go quite that small, as it might cause problems when people are trying other things on quickstart, especially since we're also pitching quickstart as a local development environment.

[csantanapr]

For CPU it will consume ~0.3 CPU on idle after they finish the tutorial, they can continue using it for other things they will scale to zero. also we are not setting a limit, they will be the ones setting the limits they want to use by adjusting the VM of docker for kind and and minikube config set

[csantanapr]

TLDR; removing the requests is orthogonal to their ability to run other things in their local cluster

[csantanapr]

Another issue I found today is that we create the VM in minikube with a flag of 3 CPU and there is no way to workaround.

Using kn quickstart minikube

😄  [knative] minikube v1.25.1 on Darwin 12.1
✨  Using the docker driver based on user configuration
- Ensure your docker daemon has access to enough CPU/memory resources.
- Docs https://docs.docker.com/docker-for-mac/#resources

⛔  Exiting due to RSRC_INSUFFICIENT_CORES: Requested cpu count 3 is greater than the available cpus of 2

[github-actions[bot]]

This issue is stale because it has been open for 90 days with no activity. It will automatically close after 30 more days of inactivity. Reopen the issue with /reopen. Mark the issue as fresh by adding the comment /remove-lifecycle stale.

[psschwei]

/lifecycle frozen

[psschwei]

We've added a page to the docs specifying a minimum of 3 CPUs and 3 GB of RAM for a quickstart cluster.