andrijapanicsb commented on issue #3600: Allow deploy Admin/System VMs in disabled zones/pods/clusters URL: https://github.com/apache/cloudstack/pull/3600#issuecomment-588483080 @GabrielBrascher I gave this another test, where I had 2 clusters of 2 KVM hosts. - admin VMs - all fine - VRs - all fine - CPVM/SSVM not fine as below, similar/multiple issues - the behaviour is identical as before this PR ### test disabled clusters only - settings "allow.system.vm.on.disabled.resources" by default is set to false (please update the description of PR - it is NOT set to "true" for systemVMs) - Disabled both clusters (not zone/pod/host) - stop or destroy SSVM - and it still gets started/new one created and started even though both clusters are disabled. ### test disabled pod - same setup as when disabling the cluster, just now we are disabling the POD only - identical behaviour for SSVM/CPVM ### test disable zone (no changes to behaviour from before) - same setup as when disabling the clusters, but now we are disabling the zone only - a stopped SSVM will NOT be started autoamtically, but can be started manually (this si the same as before, I've briefly tested/confirmed) - a destroyed SSVM will NOT be recreated - so this is fine (as before) ### test disable zone, but allow.system.vm.on.disabled.resources=TRUE - disable the zone - identical behaviour as when allow.system.vm.on.disabled.resources=FALSE - i.e. as before this PR So this is not working for SSVM/CPVM definitively - if you don't want to extend the scope to cover CPVM/SSVM properly (which I'm fine with actually...) then perhaps confirm this is in line with your code (what I tested here) and rename just the setting to "...vr...." instead of "..system.vm....." - as it only affects VR per my testing. One thing to keep in mind, if only one cluster is disabled, the VM can be always started in another cluster irrelevant of these 2 settings you introduced here (funny enough - the volume is **silently migrated** to the primary storage pool of that other cluster and VM is started now in the other cluster which is still enabled) - so this is not a valid test - you need to disable all resources (i.e. all clusters, like I did in order to see if it works as expected) Thoughts?
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
