Hey everyone,
just wanted to check in before actually attempting to create a feature request in bug zilla.
I am already using proxmox since 2018 and i love the work that's put into this great project.
When heavily utilizing the proxmox api in a bad way e.g. newly authenticating for every request sent, the proxmox daemon gets a little bit unstable.
That's something that i sorted out some time ago but legacy stuff keeps us still back from reimplementing...
But today i noticed something very interesting
I was working on a proxmox cluster and trying to migrate all of the CEPH Disks in to another external CEPH Cluster, that works without any issues.
But i noticed the old RBD was still existing, so i checked why and found out that there are still active watchers on the RBD image which has to be deleted.
As the VM was shutdown anyway i thought - wait - that can't be possible right? Maybe there is a "stale" mapped rbd because of networking or io issues.
But then i discovered the watcher is active on another node where the vm config doesn't reside, so i migrated the vm to that node and could instantly observe the uptime of that vm in the gui, crazy isn't it!
Actually i would have expected the move disk task to fail as the deletion of the to be migrated RBD disk failed, it didn't even show up as unused disk.
This is something that i noticed a lot when implementing the API, the responses are very minimal.
When deleting a VM and cloning it again sometimes the cloud init config drive isn't deleted altough the VM Delete task didn't fail.
This causes a new Clone to fail as the config drive already exists on the destination storage.
All of that just made me wondering if those issues have been observed by others or the great proxmox developers.
Do we only observe this kind of issues because of heavily utilizing the proxmox api?
Would it make sense to put a little more focus in detecting "ghost" vm processes, vm disks and config drives?
Should i create a feature request to at least address the issue of displaing not deleted ressources on the source storage when moving disks?
Thanks for reading, happy to hear about your obsaminations.
just wanted to check in before actually attempting to create a feature request in bug zilla.
I am already using proxmox since 2018 and i love the work that's put into this great project.
When heavily utilizing the proxmox api in a bad way e.g. newly authenticating for every request sent, the proxmox daemon gets a little bit unstable.
That's something that i sorted out some time ago but legacy stuff keeps us still back from reimplementing...
But today i noticed something very interesting
I was working on a proxmox cluster and trying to migrate all of the CEPH Disks in to another external CEPH Cluster, that works without any issues.
But i noticed the old RBD was still existing, so i checked why and found out that there are still active watchers on the RBD image which has to be deleted.
As the VM was shutdown anyway i thought - wait - that can't be possible right? Maybe there is a "stale" mapped rbd because of networking or io issues.
But then i discovered the watcher is active on another node where the vm config doesn't reside, so i migrated the vm to that node and could instantly observe the uptime of that vm in the gui, crazy isn't it!
Actually i would have expected the move disk task to fail as the deletion of the to be migrated RBD disk failed, it didn't even show up as unused disk.
This is something that i noticed a lot when implementing the API, the responses are very minimal.
When deleting a VM and cloning it again sometimes the cloud init config drive isn't deleted altough the VM Delete task didn't fail.
This causes a new Clone to fail as the config drive already exists on the destination storage.
All of that just made me wondering if those issues have been observed by others or the great proxmox developers.
Do we only observe this kind of issues because of heavily utilizing the proxmox api?
Would it make sense to put a little more focus in detecting "ghost" vm processes, vm disks and config drives?
Should i create a feature request to at least address the issue of displaing not deleted ressources on the source storage when moving disks?
Thanks for reading, happy to hear about your obsaminations.
