Best practice for running reclaim

Incase you are running zero reclaim on devices from the same thin pool.. they will be queued up. It runs on multiple devices at the same time but also it can't run on all the devices simulataneously.

Of course this also depends on how many different pools you have, directors etc. Start smaller and scale up.

This document might help. More application focused than you may want but should get you started

http://www.emc.com/collateral/software/white-papers/h6730-virtual-provisioning-space-reclamation-wp.pdf

I believe there would be some performance impact.

Running the space reclamation command will spawn a DA background task that will examine the allocated extents on specified thin device. For each allocated extent, all 12 tracks will be brought into cache and scanned to see if they contain all zero data. If the entire extent contains all zero data, the extent will be de-allocated and added back into the pool, making it available for a new extent allocation operation.

Thanks & Regards,

Arjun Singh

Well zero reclaim is a low-priority task so the DA will throttle the reclaim rate, possibly significantly, in the face of activities with higher importance such as host I/O. You may not see much of a performance impact as it has been designed not to. That being said it does still consume resources so there are limits to what throttling can do, especially if your DAs are busy. I wouldnt reclaim in the 100s of devices at once, but in the tens is probably fine.

With virtual provisioning I would expect each device to be on all disks and BE directors. Does this mean that the reclaim will only run against one device at a time?

That is a helpful document but I am not seeing where it discusses the performance impact on the VMAX during the reclaim. For example if I have 400 devices to run a reclaim against can I just kick them all off and the VMAX will throttle it or ....? Thank you.

Hello,

The following document should help you with your queries :

https://support.emc.com/docu32785_White-Paper:-EMC-Symmetrix-VMAX-Virtual-Provisioning-Space-Reclamation-and-Application-Considerations.pdf

Thanks & Regards,

Arjun Singh

There will be performance impact definitely. Runnning zero reclaim will keep the backend directors busy. IF the devices are from the same disks and BE directors , then the devices will be queued.

Also you need to suspend the SRDF if any of the devices belong to an active session. You need to terminate the clone sessions aswell.

There wont be significant impact on the host IOs as reclaim is a low priority task for the Symm and I would recommend that you run reclaim on smaller set of devices not all 400 devices in one command

regards,

Saurabh

Hi PDX,

as far as i understood the O.S. that you are execute the PPME are probably Windows.

Where do you use the OR? For UNIX and in general for Open Systems?

There are tools for Windows that you can reclaim the space and for open systems you can the DD if i remember well to do the reclamation.

Also it depends the Volume Manager that the O.S. are using.

You can do reclaim on an ESX host from 5.0 and latest i mean 5.1.

On Windows platforms from Windows 2008 and on Open systems that are using Veritas from 5.0.1 MP1 if i rememeber well.

So you have to give us more info to help you please.

From the Storage side, reclaim can be done at any time and as the Sumanthn and Saurabh said you must execute reclaim for a small set of devices (you are kindly requested to forget the number 400 devices) and you won't have any performance issue. Try to execute the reclamation non-working hours.