V-Max Best Practice FA connections

Yes, you should still pick your paths so that they are physically as far away from one another (for redundancy)

However in some cases, you may have to put the redundant path into the same engine because of configuration rules.  Every engine has to be symmetrical with regards to the type of IO modules installed.  For example, if you had a 4 engine V-Max with all fiber and 2 GigE modules, the 2 GigE modules would need to be installed in the same engine, as far as I know.

Well I don't have GigE ports all 64 are fiber.

The SRDF are also fiber.

Is it a good idea to spread 2 on even directors and 2 on odd?

If you look at the attached picture of the back of a V-Max engine, the user ports are on the bottom row (green).  One side is connected to one director the other side to the other director.   That was what I meant by opposite sides

.

You're absolutely correct, Front-End ports on the left side of the engine are the even director and right side are odd director.  As far as configuring for any V-Max system, for best availability you should try to do opposite director on another engine strategy for your multi-pathing planning.  For example, if you have a 2 port host, place one host port on an odd director port, say 7f port 0, then the other port on an even director in another engine, say 10e port 1.  If you have a 4 port host, then attempt to put one port on each engine, and have two on odd directors, and two on even directors.

What you should try to avoid as much as possible is placing the multi-pathing ports on the same engine and/or director board.  Of course, if you have a regular V-Max with just a single engine or V-Max SE model, you can't use jump engines as a strategy, as you only have one engine, so then you go with opposite director.

Hope this helps,

PSE Chris

Hello all, I have a follow-up question to this thread... 

Other than physical separation of fiber connections is there any other reason to pair FA's across engines?  For even numbers of engines it's a no-brainer to follow rule of 17.  However I have several VMAX's being deployed with an odd number of engines.  From all of the documentation I have read the engine itself is a fault tolerant component (otherwise who'd buy a V-MAX SE?).  A single engine carries both DA pairs for a quadrant of disks, and if you lose the engine itself you essentially lose connectivity to up to 360 drives so host connectivity is the least of your worries.

Thanks,

You dont lose access to storage if an engine goes down, all storage is available to all engines. You will lose access to configure the array, each engine controls one section of disk and all engines need to be available in order to configure it. This was validated when another customer I know failed an engine a few days back.

As posted earler, you should not have all connections from a server on one engine. If you do your server will lose access to the storage in case of failure or selective upgrades. I was told upgrading cache on an engine requires that engine to be offline. 

An EMC Solutions Architect basically said the same thing. However my CE about comes to the same conclusion I do, which is you are dead in the water if an engine dies. This is my understanding of DA connectivity, using a 2 engine 2 storage bay VMAX as an example:

- Engine 4's DA's connect to the lower 240 drives in bays 1B and 2B.

- Engine 5's DA's connect to the upper 240 drives in bays 1B and 2B.

If engine 4 dies, how does engine 5 get to the disks in the lower half to handle host I/O? I haven't found any documentation that describes this scenario.

That doesn't work with RAID-5 or RAID-6. Too many spindles involved.

I know another customer that lost an engine and all of their storage stayed available. Even during the replacement the storage was still available.

The drives in that engine must have been protected by drives in another engine, or by devices over RDF.

Think of an engine as a DA pair.  So if in a DMX2500 both DAs of a pair fail, we would lose access all access to all those drives, and in a DMX1500 if both DA boards fail, we lose all access to all drives.

Both directors in an engine are completely independent.  You could say the engine is equivalent to a chassis in a DMX3/4.

That's the analogy I've heard as well. But what I am hearing (and reading) is an engine can go offline with no impact. I disagree with that statement because no one has been able to tell me how a non-failed engine can access the drives behind a failed engine's DAs.

How does a multi-engine V-MAX stay running with a failed engine?