[Wishlist] datastore maintenance and categories

Just a short post on some thoughts I had on improvements for VMware vSphere when I wanted to delete a datastore holding 10 VMs.

Wouldn’t it be great to be able to put a datastore in maintenance mode like you can with an ESX host. When the datastore is set into maintenance mode, vCenter would Storage VMotion the VMs to other datastores and also make sure that other data stored on it, would be moved, the dvSwitch info for example. If this was possible it would greatly decrease the time an admin needs to clean out a datastore.

Pro’s:
– Admin doesn’t have to wait for each VM to SVMotion, he can just set and forget. When I now want to SVMotion all those VMs from the datastore I have to run through the wizard for each VM.
– Admin doesn’t have to think about VM placement. Searching all the other datastores for enough room, vCenter will figure it out.
– Datastore in maintenance mode would “lock” the datastore for new VMs, preventing other admins to place new VMs on it since that datastore had so much free space while you were just in the process of cleaning it.

Con’s:
– VMs are often placed on certain datastores because performance characteristics between datastores may vary. Some have RAID5 below, some have SATA, some have fiberchannel, some have SSD, etc, etc. vCenter wouldn’t know about this and just move VMs. Therefore I thought of a second whish:

Datastores can have a category label. This category will give you the option to place datastores with the same performance characteristics together. vCenter would then only SVMotion between the same category, unless manually overruled of course :-)

Let me hear your thoughts on this in the comments.

9 thoughts on “[Wishlist] datastore maintenance and categories

  1. This makes a lot of sense, Gabe. In fact, it makes we wonder why something like this isn't already implemented. I'm sure you could emulate the datastore maintenance mode feature with some PowerCLI magic, but this should be implemented in the VIC and not rely on scripting skills. We're not Alan or Arne, at least not all of us. :)

  2. 1. A good 1st step might be to just freeze the datastore for any “new” VM's so you can remove it safely later.

    2. Another idea might be to have a standard “method” within vmware to signal your intent to delete a Datastore and have vmware not allow a removal if it still has files/entries present on the datastore.

    I think this idea is worth thinking on some more…

  3. Gabrie,

    This is a great idea. You are right in that movement of VMs would need to be carefully examined for performance characteristics, among a few other things that come to my mind.

  4. This feature would be really cool but I think the idea of categories is absolutely necessary! Maybe you have business critical VM's in a dedicated datastore which are replicated to your disaster recovery site. If these VM's are then migrated to a datastore without replication and you loose your primary site you'll probably have a problem. ;)

  5. I would like this. I would also like it that you can do fault tolerance on storage level (=effectivily a raid mirror over multiple luns). In my eyes, san is still the weak spot in a vm solution.

Comments are closed.