Striped Meta Expansion / Clones / BCVs ?

I may misunderstand your question. Here are few thoughts:

>>> it can expand by any amount very fluidly

It can be done in VMAX as well with concatenate as meta configuration. However as you said its not true for stripe meta - as data needs to be redistributed when adding a new stripe (when data integrity/preservation is requested)

>>> Is it possible to create a larger LUN, then copy the data from a smaller LUN in there, and then present that LUN to the host?

While expanding stripe meta, if one selects preserve data option then the data should be preserved (using BCV). Personally, Symm already does this efficiently, i don't think you will save any time by doing it manually. i may be wrong - but that's my expectation.

Btw, what is the main concern? is it regarding the time which symm is taking while expanding stripe meta (with preserving data)? Have you considered automating these using scripts - you can automate these steps either by SymmCLI cmds or ESI PowerShell Toolkit Cmdlets.

Hope this helps!

Regards.

Hi rwise,

Welcome to EMC Community.

I tottaly understand the situation you are in. Expansion of devices on the fly is not supported with the current generations of VMAX (It is supported with other arrays like EMC VNX & CX). This feature will be addressed and supported in next generation of VMAX (VMAX V3).  Regarding your question about the clones, You can clone from a smaller device to a larger device however, you cannot restore data from the target device and bear in mind that you will be limited also to max device size on VMAX40K which is 240 GB so if you are going beyond this, then META devices should be considered.

Another restrictions we have for cloning metas is that source and target devices should have same configurations which means that:

Concatenated meta can only be cloned to concatenated meta and Striped meta can only be clone to striped meta with the same size/number of meta members and also same stripe size.

The only possible solution I can see for your case, is to leverge host based tools like PPME  (license is for free if you have a PP license) or OM (however this is for migration not cloning).

Hope this helps.

Mohammed Salem

Hello rwise,

For your last question, Host based migration tools can help however this will use some CPU cycles from your host.

For the question regarding chargeback, There is a multitenancy support for VMAX40 with Ionix StorageScope for chargeback however, I am not sure how can we utilize this in your existing system. and I doubt this will help to override the limitation we have for striped METAs expansion.

I will ask one of my colleagues to reply to you for a better answer.

Hope this helps.

Mohammed Salem

Hi Mohammed, That would be fantastic if you or one of your colleagues could explain how to do this (i.e. simply expand a striped meta to any size arbitrarily). It shouldn't be too difficult to create a larger device/meta and copy the data from a smaller device/meta I hope, but if the only way is via a Clone license, then that would be disappointing (particularly when 5+ year old USP-V technology can do that seamlessly with ease and base functionality). I don't mind if it's complicated, I can script up anything if there is a process to get this to work. Many thanks.

i am afraid you will be disappointed, you can't expand a striped meta to any size you like.

I know I can't do that, but what about other options? Is there no posibility to create a different larger sized BCV meta and copy data in there and then convert that to a TDEV meta and present to a host for example, or some other option?

can't do that either, only options are M.Salem already mentioned, host based migration: either using PowerPath Migration Enabler or native LVM/ASM tools. Now if you are willing to take downtime there is always robocopy/rsync. Create a custom size device, present it to the operating system, shutdown application and copy the data.

Downtime is mostly ok for this client. Robocopy is great, but in my experience when we have filesystems with 100's of thousands of files, it can be very slow. but never got around to using PPME so that would be good (but unfortunately they use native multipathing on Windows 2008 / Solaris / Linux here so no PowerPath except on Windows 2003). I'd like to try out Open Migrator (but as file-based it will probabaly have the same limitation as robocopy). Are Open Migrator licenses free?

I've done 2 or 3 thousand host migrations on Open Replicator (my favourite migration tool), but I've only ever done that array to array. Long-shot here, but is it possible to OR internally between LUNs in the same array? That could work quite well (if it was possible).

i don't think OR will let you do loopback replication like SANCopy on Clariion/VNX does.

Excellent, well at least I know the limitations, thanks guys. Open Migrator does sound quite useful in fact and could be the route to go for my client to be able to migrate to an arbitrarily larger sized LUN without breaking the chargeback billing process (that is standard in every corporate / bank in which I have worked) in the way that striped meta expansion does. I'll hopefully get a chance to test this weekend. Edit: just saw dynamox's note that it requires a reboot - that almost completely negates it's use in the vast majority of the corporate environments in which I work (i.e. change management for downtime on global banking systems is not easy ...). What a huge shame ...

If anyone can think of a brilliant / ingenious other method of doing what ancient HDS technology can do in base functionality (block-based migrate / expand a LUN to another arbitrarily sized larger LUN on the same array, either offline or online), that would also be a great thing to add to my toolkit ? ...

installing OM will require a reboot

That's correct; Open Replicator is only supported between arrays (not within a given array).

Open Migrator is block-based at the host layer, so you don't have to deal with the overhead of opening/closing all the files or with preserving permissions/acls/etc... and it's free with the purchase of a VMAX. You can also fix partition alignment issues with it (if applicable), because it can migrate partitions/volumes in additional to full disks/LUNs.

Thanks,

- Sean

I don’t know if this was covered, but Open Migrator take several reboots to start and complete the migration. The data synchronization itself is an online event but you have to attach the filter driver to the volumes and before you can begin the sync and the host has to be rebooted. Once the sync is complete and you want to commit the migration you have to reboot again. This is then followed up by yet another reboot to detach the filter driver from the volumes. If you want to then uninstall OM, you have to reboot again.

Now, I’ve used OM quite a bit and it’s a good product that definitely has a place and can overcome some configuration issues that other products can’t, like a volume offset issue. The reboots are definitely a check in the negative column though.

Mike

While at this stage of the discussion, my question might seem irrelevant and out of context, but just wanted to know what auto-meta does and would that be helpful for the problem being discussed here?