Today we have a VPLEX Metro setup between arrays at two different sites. We are looking into the need to upgrade those arrays. Considering that VPLEX mobility/migration is not supported between distributed devices, how do we go about migrating from the current arrays to the new arrays using VPLEX?
What you are attempting to do is quite common, let me try to explain it further. Ill try to create a video describing this as well.
Today you have a distributed virtual volume A, which is comprised of devices B at site 1 and C at site 2. At some point the array that B is on needs to go away and you need to move B to a new array. Bring in a new array and create lun B1 that is of the same size as B and present it to vplex and claim it. You then do an extent/device mobility job from B to B1, this whole time the virtual volume A is still seen by the host. The synchronization from B to B1 is happening in the background. Any new writes to A are going to B,B1, and C.
At some point you commit the mobility job at which point now the distributed volume A is now comprised of B1 at site 1 and C at site 2.
Ok, I can see that. However, doesn't this mean that I now have two un-sync'd copies of my data, one device per site, and no virtual volume representing them to the host(s)? Syncing these together to make a new distributed volume doesn't seem to make a lot of sense, so wouldn't you instead migrate one leg of the original distributed device, and then sync it to a new LUN on the other leg using a new distributed device? If I do this I don't have a virtual volume at this point, so this would be a disruptive migration from the host point of view, right? Finally, what about in the stretch cluster scenario?
Hello Will, while its true you cannot migrate distributed Volumes you CAN migrate the devices at each site of a distributed volume. A distributed volume is made up of 2 devices/extents at each site. So its a matter of creating a device/extent on the new array, presenting it to vplex , and doing a device/extent migration.
Ok, so to continue your thinking, if you were replacing both arrays, you would then do the same with LUN C, and use a local mobility job to move to C1.
During this time the distributed volume would stay up?
ankur.patel
2 Intern
•
286 Posts
4109
1
Posted January 6th, 2015 08:00
What you are attempting to do is quite common, let me try to explain it further. Ill try to create a video describing this as well.
Today you have a distributed virtual volume A, which is comprised of devices B at site 1 and C at site 2. At some point the array that B is on needs to go away and you need to move B to a new array. Bring in a new array and create lun B1 that is of the same size as B and present it to vplex and claim it. You then do an extent/device mobility job from B to B1, this whole time the virtual volume A is still seen by the host. The synchronization from B to B1 is happening in the background. Any new writes to A are going to B,B1, and C.
At some point you commit the mobility job at which point now the distributed volume A is now comprised of B1 at site 1 and C at site 2.
make sense?