I have been doing this manually until now. We have quite a few LUNs and want to start scheduling this on each to happen regularly.
What are the pros and cons of scheduling this? What frequency should it be run?
Right now all my LUNs are provisioned in a VMware SDRS enabled cluster. We are frequently dropping below our threshold of 15% which sends out alerts. I clean and i can get ~5% back or so. Its almost time to reorganize or order more storage.
What are the pros and cons of cleaning the white space and the frequency in doing so? I have to imagine if this was a standard function, it would be be built into the XIO and do this as part of a schedule. The fact that you need the VSI plugin and it is not scheduled makes me think there is a reason not to do it regularly.
So we're doing this daily right now. For us, instead of our issue being running out of capacity, we have a wide swing in the available shared memory on the XtremIO based on the current amount of allocated blocks on our VMWare LUNs. If we let it go we get up to 94% used, if we run the unmaps we're back down to 70%ish.
Anyway, we're not going through the plugin at all, we're using the calls via the powershell powercli for vsphere (https://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=20575... and ESXcli 5.5 UNMAP and PowerCLI script? | VMware Communities for some more info) and a scheduled task on a windows machine.
It IS my understanding that ESX 6 does UNMAPs automatically for you (if you have everything set up right...vSphere 6.0 Storage Features Part 8: VAAI UNMAP changes - CormacHogan.com), so I'm hoping we'll be able to test and move to that in the next few months.
Space reclamation is not built into the XtremIO GUI itself because it is a function of the host operating system and that’s where the space reclamation command needs to be run. This has nothing to do with how XtremIO (or any other storage works). In other words, when you delete some storage from a host perspective, the host needs to send the unmap/space reclamation command to the backend storage. Some operating systems do it automatically (like Windows 2012) and some don’t (VMware).
VSI plugin is the software piece that is installed in the host (i.e. ESX server) and sends this space reclamation command per the schedule that you want to run it at. How often you run it will depend on how much deletions/VM movements are happening in your environment. Whether you run this every midnight or every two hours or once every week will depend on whats happening in your environment and how often you need to reclaim space. VSI plugin will provide this functionality. When this command is running, there is some extra workload on the array – so you would want to make sure that it is not running during the peak times of your day.
Keep in mind that VSI plugin will only run space reclamation at the datastore level. If you are doing any deletes inside the VM, you need to run space reclamation within those clients first. VSI plugin will not help with this client level space reclamation.
Is there any specific reason why you chose not to use the VSI plugin? Or you just did not know about it when you started looking at space reclamation?
We don't let our VMWare admins manage their own storage, so we don't install the VSI plugins unless there's some very large amount of value to be had. Generally the things provided by the plugins do not given you even as much as you get out of the normal array management guis. Plus who wants to be inside the Vsphere client if they can avoid it? It's a terrible thing, even the web version.
so.......how often should this be run on a schedule and are there any pros or cons to running it more or less frequently? Daily vs Weekly etc? I understand what it does (reclaim white space) and what it doesnt do (reclaim space from guest level OS's).
I think it depends on the actual status at customer side, suggest to run space reclamation after you delete any VMs or data, it will be helpful to release capacity at array side.
And when you do run space reclamation, there will be an additional workload on the array that you should be aware of (don’t run it at peak production hours, etc). How much additional workload would be there (and for how long) depends on the amount of space that needs to be reclaimed.