HOWTO - VMware-Integrated, and Application-Integrated Array Replicas

I often get asked about whether array writeable snapshots can be used to accelerate VM image deployments or as an instant backup/restore mechanism for VMs and datastores. I also get asked - can you do array snapshots with VMware-integration to make sure they are restartable images?

Simple answer... YES!

I also get the question of "why do this vs. native VMware Snapshots". I always cringe at that question - it's rarely one vs. the other - the better question is "where do I do apply this technology". ESX snapshots are very good to take to capture an individual machine state (and are used in all sorts of neat ways for that purpose - including fun examples like Lab Manager). Upside: They are lightweight to take, and lightweight to revert (both these actions involve small changes to the VMX file and changing the state of the underlying VMDK files used), Downside: they "expensive" to delete - which is the only way to reclaim space (depending where you are in the snapshot manager history, a delete "merges" the vmdks representing prior image states). Currently, you can also only do one per datastore at once, so that sequentiality causes some issues on things that depend on it (like VCB).

In general (but not always) array snapshots perform well and scale well. Every snapshot technique (every one of them) makes a deal - you pay the piper up front, or at back, or at a decoupled time from the snap, or in space consumption. That's not to imply better/worse, but more "know how they work".

But, let's do more than just answer the question... I'm a big fan of SHOW ME. Better yet, GIVE ME THE TOOLS TO TRY IT!

read on to do exactly that...

EMC Replication Manager is a tool that... well... manages replication :-) Think of it this way - every EMC array can take writeable snapshots (in spite of FUD slung by our competitors), but can also create clones, and some can apply consistency groups (many replicas handled together), some can do this across arrays (only the DMX), and even journalled recovery points (if you think of snapshots/clones as a "point in time copy", a journal is like a Tivo). That's a lot of choice, and a lot of flexibility - a hallmark we try to maintain. There's a solution fit for almost any use case. If you see a theme in my posts - there is **no one answer**, and that it's the know-how more than anything else that makes IT tick.

Replication Manager (RM) is:

a single pane of glass that can manage almost all EMC's replication technologies (and any place where we have replication it doesn't integrate with is quickly closed)
Integrates into the use case - ergo (and this is a small set):
- for Exchange replicas - used for backup offload/instant mailbox and database restore, RM manages the VSS and log truncation/handling
- for SQL Server/Oracle replicas, - used for rapid test/development copies, instant restore and backup offload, RM manages the VDI/VSS/Oracle Hotbackup mode and log handling;
- for VMware environments - used for instant VM or datastore-level restore - integrates with vCenter to map volumes and storage, and also ensure VM quiesce for restartable copies.
Has advanced scheduling and user authentication - so you can have people use it that may not be authorized to manage the array - for example VMware administrators or DBAs, or Exchange Admins.

The GUI looks like this:

Here's a quick demo of the tool in action - by another one of our VMware Technical Specialists - Aaron Chaisson

The current release (RM 5.2) is very simple, very mature, and very stable in my experience, having gone through the early days of RM 2.x and 3.x. We've been shipping this for a long time, and in fact, were the first vendor to ship the VMware Integration, though didn't market it well enough (we called it RM 5.12, rather than RM for VMware :-)

NetApp has a similar tool called SnapManager for Virtual Infrastructure part of the SnapManager family, and Dell/EqualLogic has something similar called Auto Snapshot Manager/VMware Edition (as far as I know, these are the only examples that do this - anyone else out there, if I'm incorrect, please point them out).

These tools do something in common - before the array snap, optionally they invoke the ESX snap to ensure that the VM can be restarted - otherwise its a "crash consistent" replica. "Crash consistent" just means it's in a "non clean shudown state". Not terrible, but not great. I know this is an optional setting in RM, and believe that to be the case with SMVI, though am not sure (any NetApp folks want to confirm?). The reason it's optional is that it can make the job take long if you have datastores with a very large number of VMs (as each VM is snapshotted in sequence). The deletions aren't "expensive" as they occur right after the snap, so divergence is small. We're working with VMware to develop another way to quiesce VMs.

But - hey, don't listen to me! Try it for yourself! If you're an EMC employee, an EMC partner, or most of all, an EMC customer - you can download a version and use it for 90 days. It works with the Celerra VSA (here), so if you want to try it in a sand-box, go to town! I WANT TO KNOW YOUR THOUGHTS!

My Swiss colleague, Marcel Brunner has done a video series on how to install it and get it up and running. Marcel - THANK YOU!

Part 1: Downloading Replication Manager