Skip to content

Wrapper for installing managed clusters on the available platforms

License

Notifications You must be signed in to change notification settings

cloud-bulldozer/hcp-burner

Repository files navigation

title
HCP Burner Docs

What is HCP-Burner

hcp-burner is a wrapper to automate create-use-destroy of OCP clusters in the different managed platforms.

How HCP-Burner works

Following helpers are available for every platform:

  • Schedulers:
    • Install
    • Execute Workloads
    • Cleanup

Every platform needs to overload following functions to be integrated in the cluster:

  • Initialize
  • Create Cluster
  • Delete Cluster
  • Platform Cleanup
  • Watcher

Available Platforms

As mentioned on Red Hat Cloud Services, The foundation of Red Hat Cloud Services is Red Hat OpenShift®, a comprehensive application platform. In addition to self-managed offerings, Red Hat OpenShift is available as a cloud service directly from Red Hat as well as from major cloud providers.

This wrapper will be adapted to run on every platform where Red Hat Openshift will be available as Cloud Service. On that purpose, we have establish a tree of platform - subplatform structure to fill that requirement. A platform will match every cloud provider and a subplatform will be an installation method of Openshift on that cloud provider.

Following platforms are available at this moment:

Arguments, parameters and configuration options

Almost all the parameters can be defined in three ways:

  • Wrapper parameters: hcp-burner.py --cluster-count 100
  • Environment Variable: HCP_BURNER_CLUSTER_COUNT=100 hcp-burner.py
  • Configuration File: hcp-burner.py --config-file ./hcp-burner.ini

Only parameters --platform and --subplatform must be defined as wrapper arguments, platform is always required but subplatform is optional

To add any other parameter to the config file, remove -- from the argument and change _ to _, for example:

--cluster-name-seed will be:

[Defaults]
cluster_name_seed = test

Full version of a config file can be found on hcp-burner.ini file

Preference

All parameters will have following preference when they will be defined in more than one place:

Argument > Environment Variable > Config File

Common arguments

To use the config file, define common parameters under the [Defaults] section

Argument Default Value Config file variable Environment Variable
--config-file HCP_BURNER_CONFIG_FILE
--install-clusters
--platform platform HCP_BURNER_PLATFORM
--subplatform subplatform HCP_BURNER_SUBPLATFORM
--uuid HCP_BURNER_UUID
--path HCP_BURNER_PATH
--cluster-name-seed HCP_BURNER_CLUSTER_NAME_SEED
--static-cluster-name HCP_BURNER_STATIC_CLUSTER_NAME
--workers 3 HCP_BURNER_WORKERS
--workers-wait-time 60 HCP_BURNER_WORKERS_WAIT_TIME
--wait-for-workers
--cluster-count 1 HCP_BURNER_CLUSTER_COUNT
--delay-between-batch 60 HCP_BURNER_DELAY_BETWEEN_BATCH
--batch-size 0 HCP_BURNER_BATCH_SIZE
--watcher-delay 60 HCP_BURNER_WATCHER_DELAY
--wildcard-options HCP_BURNER_WILDCARD_OPTIONS
--enable-workload
--workload-repo https://github.com/cloud-bulldozer/e2e-benchmarking.git workload_repo HCP_BURNER_WORKLOAD_REPO
--workload cluster-density-ms workload HCP_BURNER_WORKLOAD
--workload-script-path workloads/kube-burner-ocp-wrapper workload_script_path HCP_BURNER_WORKLOAD_SCRIPT_PATH
--workload-executor /usr/bin/kube-burner workload_executor HCP_BURNER_WORKLOAD_EXECUTOR
--workload-duration 1h HCP_BURNER_WORKLOAD_DURATION
--workload-jobs 10 HCP_BURNER_WORKLOAD_JOBS
--cleanup-clusters
--wait-before-cleanup 0 HCP_BURNER_WAIT_BEFORE_CLEANUP
--delay-between-cleanup 0 HCP_BURNER_DELAY_BETWEEN_CLEANUP

ElasticSearch arguments

To use the config file, define common parameters under the [Elasticsearch] section

Argument Default Value Config file variable Environment Variable
--es-url HCP_BURNER_ES_URL
--es-index hcp-burner HCP_BURNER_ES_INDEX
--es-index-retry 5 HCP_BURNER_ES_INDEX_RETRY
--es-insecure

Logging arguments

To use the config file, define common parameters under the [Logging] section

Argument Default Value Config file variable Environment Variable
--log-level INFO HCP_BURNER_LOG_LEVEL
--log-file HCP_BURNER_LOG_FILE

About

Wrapper for installing managed clusters on the available platforms

Resources

License

Code of conduct

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published