Skip to content

Latest commit

 

History

History
173 lines (141 loc) · 7.47 KB

README.md

File metadata and controls

173 lines (141 loc) · 7.47 KB
Raw

Hibernate your Hive provisioned clusters and scale cluster pools

This repository has the source code as well as the deployment for two Kubernetes CronJobs that will Hibernate and Resume your OpenShift clusters.

In addition, there is tooling to create CronJobs for scaling cluster pools.

How this repository can be used to hibernate Hive provisioned clusters

There are three supported configurations you can use:

  1. Opt IN: A cluster will only be affected by the cronjob if it has the label hibernate: true
  2. Opt OUT: A cluster will ALWAYS be affected by the cronjob, unless it has the label hibernate: skip
  3. Cluster Scoped: The cluster in the namespace where the cronjob is created is the only one affected by the ACTION value you choose (hibernating or running)

Requirements

  • Red Hat Advanced Cluster Management for Kubernetes v2.1+
  • OpenShift oc command line tool installed
  • OpenShift 4.5+ clusters provisioned by ACM
  • Add the NAMEPSPACE parameter, pointing to where you want to run the hibernation jobs, in the ./options.env file (Opt IN/OUT)
# ./options.env
NAMESPACE: open-cluster-management  #This can be any namespace where you want to install for Opt IN or Opt OUT

Deploy Opt IN or Opt OUT scenarios

  1. Log into your ACM Hub on OpenShift
  2. Run the command: 
    make cronjobs
  3. A ServiceAccount called hibernator was created. To grant the required privileges cluster-wide, run the command: 
    make roles
    If you prefer to grant finer-grained privileges for specific namespaces, see templates/roles.yaml for details.
  4. Monitor the CronJobs 
    oc get cronjobs

Deploy cluster scoped scenario

  1. Log into your ACM Hub on OpenShift
  2. Switch to the namespace(project) of the cluster you want to apply the cronjob to
  3. Run the command to create cronjob 
    # ACTION=hibernating   or  ACTION=running
# SCHEDULE=<crontab_format>
# NAMESPACE=<my-cluster-name>  The cluster name you wish to apply the cronjob to

# Hibernating cronjob
oc process -f templates/cluster-scoped-cronjob.yaml ACTION=hibernating SCHEDULE="0 23 * * 0-6" NAMESPACE= my-cluster-name --ignore-unknown-parameters=true  | oc apply -f -

# Running cronjob
oc process -f templates/cluster-scoped-cronjob.yaml ACTION=running SCHEDULE="0 12 * * 0-6" NAMESPACE= my-cluster-name --ignore-unknown-parameters=true  | oc apply -f -
    _Note: The cluster scoped cronjob requires the value hibernating or running to be lowercase. This is becuase the ACTION parameter is also used to generate the cronjob name.

Customizations for Opt IN & OUT

For a full list of options, run:

make params

To change any options, create or edit the options.env file. (NOTE: This file is not under version control.) For example, to disable automatic resumption of clusters, add the following to the options.env file:

RUNNING_DISABLED=true

Time window

The default setting is to have the clusters running from 8am - 6pm Mon - Fri EST. You can adjust the times by adding the HIBERNATING_SCHEDULE AND RUNNING_SCHEDULE params to options.env.

It uses the standard CronJob format and the check is done against a UTC clock. For EDT, that means +4hrs.

#       Time is measured in coordinated universal time (UTC)
#                ┌───────────── minute (0 - 59)
#                │ ┌───────────── hour (0 - 23)
#                │ │ ┌───────────── day of the month (1 - 31)
#                │ │ │ ┌───────────── month (1 - 12)
#                │ │ │ │ ┌───────────── day of the week (0 - 6) (Sunday - Saturday)
RUNNING_SCHEDULE=0 13 * * 1-5"

Skipping Clusters

The job will only target clusters deployed by Hive. It looks for the ClusterDeployment objects. If you put a label on these objects hibernate: skip they will be ignored by both CronJobs.

For an opt-in model, instead specify OPT_IN=true in the options.env file. In this case, only ClusterDeployments with the label hibernate: 'true' will be acted upon.

Run once job

You can hibernate or resume all applicable clusters manually. These commands use the options.env file, so the value of the OPT_IN parameter is respected.

To bring clusters back to "Ready"

make running

To hibernate clusters

make hibernate

Manual updates

Edit the ClusterDeployment resource for the cluster you want to change the state for. Find the spec.powerState key and change the value to either: Hibernating or Running

Building yourself

You'll need docker and a connection to a registry that your OpenShift can reach. Export the environment variables REPO_URL with your own registry URL and VERSION the tag. See make all for the development related targets. After pushing an image, you can use your own image by setting the CURATOR_IMAGE parameter in options.env.

Running local

Setup the following environment variables, then run go run ./pkg

export REPO_URL=             # The repository you will upload your image too
export VERSION=              # The image version you want to use

Time chart

Time Zone UTC offset Example
EST +5hr 11:22EST = 16:22UTC
EDT +4hr 11:22EDT = 15:22UTC
PST +8hr 11:22PST = 19:22UTC

Notes

  • Added to the open-cluster-management community 2020-11-25

Cluster pool scaling

Use Kubernetes CronJobs to scale up or down cluster pools at specific times. Reducing the size of a cluster pool during off-peak hours can help reduce cloud expense.

Customizations

For a full list of options, run:

make clusterpool-params

NOTE: This feature uses the ServiceAccount called hibernator that was created previously in the section above.

For example, to scale up a cluster pool named aws-4-10, create a file aws-4-10-scaleup.env containing the following options. (NOTE: This file is not under version control.) For example, to change the pool size at 8am Mon - Fri EST:

CRONJOB_SCHEDULE: 0 12 * * 1-5
SERVICE_ACCOUNT_NAME: hibernator
CLUSTER_POOL_NAME: aws-4-10
CLUSTER_POOL_SIZE: 4
NAME: aws-4-10-scaleup
NAMESPACE: my-namespace

For a complementary scale down of a cluster pool, create a file aws-4-10-scaledown.env containing the following options. (NOTE: This file is not under version control.) For example, to change the pool size at midnight Tue - Sat EST:

CRONJOB_SCHEDULE: 0 4 * * 2-6
SERVICE_ACCOUNT_NAME: hibernator
CLUSTER_POOL_NAME: aws-4-10
CLUSTER_POOL_SIZE: 2
NAME: aws-4-10-scaledown
NAMESPACE: my-namespace

Deploy the CronJob

Run the following command to create the CronJob to sclae up the cluster pool:

oc process -f templates/scale-clusterpool-job.yaml --param-file ./aws-4-10-scaleup.env  --ignore-unknown-parameters=true | oc apply -f -

Run the following command to create the CronJob to sclae down the cluster pool:

oc process -f templates/scale-clusterpool-job.yaml --param-file ./aws-4-10-scaledown.env  --ignore-unknown-parameters=true | oc apply -f -

View CronJobs

  1. Log into your ACM Hub on OpenShift
  2. Monitor the CronJobs 
    oc get cronjobs -n <NAMESPACE>

Manual updates

Edit the ClusterPool resource for the cluster you want to change the size for. Find the spec.size key and change the value to the size you want.