OSD 4 platform evaluation for Eclipse Che

[ibuziuk]

Description:
The goal is to evaluate OSD v4 as a platform to run Che workspaces

• Are the performances enough for our dogfooding (in terms of network, storage, compute resources).
  yes, it is enough for doing che-in-che
  build che in che works just fine against ebs gp2 volume
  • build without tests

[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 01:59 min
[INFO] Finished at: 2020-04-10T11:14:55Z
[INFO] Final Memory: 302M/1884M
[INFO] ------------------------------------------------------------------------

• build with tests

[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 05:22 min
[INFO] Finished at: 2020-04-10T11:21:50Z
[INFO] Final Memory: 243M/1967M
[INFO] ------------------------------------------------------------------------

• What are the costs?

• https://www.openshift.com/products/dedicated/ (Single availability-zone vs. Multiple availability-zone)

• How much are we going to be able configure the Kubernetes cluster?

  • dedicated-admin is the highest permissions that could be obtained - https://docs.openshift.com/dedicated/4/administering_a_cluster/dedicated-admin-role.html#dedicated-administrator-role
  • it is possible to manage sa, quotas and limit ranges
  • it is NOT possible to install privileged and custom operators from operatorhub\
  • it is NOT possible to create ClusterRoles (Main reason why arbitrary opertors can not be deployed on OSD)

• it is NOT possible to create arbitary CRDs

• it is possible to install curated operators to the openshift-operators namespace - https://docs.openshift.com/dedicated/4/operators/olm-adding-operators-to-cluster.html#olm-adding-operators-to-a-cluster

• Is it possible to build containers from a workspace with buildah there.

  • Does not work OOTB failing with - there might not be enough IDs available in the namespace
    in general buildah does not work properly in non-privileged containers

• How stable is it?

  • Engineered, operated, and supported by Red Hat with a 99.9% uptime SLA and 24x7 coverage.

• How often does it get updated.

  • The update cycle is still under discussion but in the general case, updates happen quite often - in the general case customers are supposed to get OSD updated in less than a month after the OpenShift release (z-stream / minor)
    NOTE: Updates could be put on hold if a blocker is found

• Where will be the nodes located geographically?

• A customer can choose either a Single-AZ or Multi-AZ (stretch) deployment for each OpenShift Dedicated cluster. In addition, customers can choose from standard, memory-optimized, or compute-optimized application nodes. Cluster availability region, cluster id, VPC peering/VPN configuration, and user authentication integration are also customizable.

• All AWS regions are available except China and GovCloud. Multi-AZ clusters can only be deployed in AWS regions with 3+ availability zones.

• How hard is it to deploy/update Che there?

  • Atm, it is possible to install only via oc apply & custom scripts
    • operatorhub is not supported since Eclispe Che is not part of the curated operators.
    • chectl require creation of arbitary CRDs which is not allowed on OSD
    • deploy script was removed in the upstream in 7.11.0 - Remove deploy.sh eclipse-che/che#16376

[ibuziuk]

Closing as done