OpenShift
OpenShift copied to clipboard
Azure
Start/Stop cluster
Cut down cluster costs by starting/stopping cluster while maintaining cluster state and configuration. Ability to scale down to zero-node ARO clusters.
Bump - Is there an ETA for this and is there an interim recommendation for cutting down costs please?
So did this ever get worked out? Still paying a bunch for test clusters that could be turned off when not in use.
Priority: Not only cutting costs, but refreshing VM state to recover from errors or force-update things like DNS.
Cluster restart is supported by OpenShift for every platform other than Azure-integration. (cf https://www.redhat.com/en/blog/how-stop-and-start-production-openshift-cluster as well as other Red Hat OpenShift docs which, if believed and followed, can actually lead to destruction of a running ARO cluster by shutting down the machines)
It's supported by Microsoft for non-OpenShift (Kubernetes) clusters. (see az aks stop)
The only thing that prevents restarting masters, then workers (as I see it) is a deliberate security policy decision to instantiate a read-only DenyAssignment that blocks VM control, with the only exception being given to an internal enterprise app for cluster management (which app seems undocumented and exposes no administrative functions). Simply removing this DenyAssignment would allow an admin to restart the VMs. Perhaps Microsoft Azure team considers this "too dangerous and undocumentable" at this time? Yet this can't be "rocket science research" - as every other cluster seems to allow restarting, and it MUST be done when a new ARO cluster has just been provisioned and is started for the first time? (My point is, this is "beyond the pale", my team could not believe it, so Please consider making it highest priority, as in why would people choose Azure OpenShift when it has this glaring omission of a basic function with nothing done about it for years?)
Related past support requests: https://learn.microsoft.com/en-us/answers/questions/448680/how-to-remove-deny-assignment-from-aro-cluster-whi.html https://learn.microsoft.com/en-us/answers/questions/681940/index.html and, indirectly: https://github.com/Azure/OpenShift/issues/217
No idea as to why this is handled so unprofessionally by Azure. This is a glaring omission and lack of functionality. Touting all the benefits of Openshift and then not being able to correctly shutdown and start a cluster is really bad behavior.
How can we get more attention to this feature? This feature request is in the backlog for 2 years now. Either close it or work on it. Backlog items for 2 years without any progress or additional info do not help anyone.
This is an absolute trap that openshift clusters cannot be stopped , paying for QA clusters running during nights and weekends
September 2023, the missing functionality, which incurs heavy costs - basically making the product unusable for qa/test, was reported more than 2.5 years ago! An ETA was asked in each year ... are you guys for real? you're selling this product https://azure.microsoft.com/en-us/pricing/details/openshift/ ... what's the fuss in having a 'shutdown cluster' option????
Any updates on this? Is this on the horizon or we can assume that this won't be done?
FYI we've mostly (not entirely) migrated to replicated for our use-case. We needed the Open shift clusters mostly to run integration tests and setting up and tearing down an ARO cluster could take hours so we left them running because this feature doesn't exist. In reality we only need them a tiny fraction of the time.
I do understand that this is not possible for every use-case. For ours it works perfectly fine.
https://www.replicated.com/compatibility-matrix
