«

»

Dec 01

FC@Home – Redux….

The rack before the upgrade - note no drive array.

Ok, it’s amazing what can goes on when you finally get some time at home to do the things you’ve been needing to do for centuries.

Several months ago I bought an old Dell PowerVault 660F.  This is a simple 14 drive array, with two Mylex FFA Based controllers (despite what Dell would have you believe)

I have had *NOTHING* but trouble getting it working.  Mostly because of course Dell stopped supporting the Mylex based arrays right around the time it started getting some “real” enterprise class hardware.  (A-la EMC).

So thanks to a customer who must have *REALLY* liked me, I got a new 42U rack and a couple of Brocade switches to play with.  That plus a $250.00 investment in the old PowerVault 660F, and I have a SAN.

So a couple of weekends ago I was home alone, my wife off doing girl-stuff with her best friend for the weekend (the best present I could have given her – a weekend off.)

The two systems required to keep this (and other) sites running are moved to a bakers rack in the corner.

I took this old Compaq rack and started tinkering with it.  First I consolidated all of my VM systems to one host (Dell 2650, 2CPU, 6GB Ram) and put it on a shelf in the corner.  Then I stacked the Windows system that is my AD environment (and with over 2TB of disks in it, my Vertias NBU environment and fileserver).  This enabled me to shut the rest of the environment down.

I then proceeded to gut it.  Took the empty compaq rack and started loading it with dell hardware.  (Odd combination to say the least, but beggers can’t be choosers)

I found some rapid rails for Dell 2650’s on Ebay for $12.50/each, and one peice at a time moved everything into the new rack, taking great care to be on my best behavior cable-management wise.

I had help during this process...

As it all came together, batteries in the bottom, 5 servers, Keyboard/Monitor tray – (cool one with the 8 port KVM switch in the back of it) then Array, Switches, Tape Library, PDU stuff and a handy slide-out monitor for the top of the rack.)

Took me about 2 days to really get it together.  And got it moved into place before the wife returned from her holiday. 🙂

Now back to what started this.  I have spent the ensuing two weeks working and job hunting, so not a lot of time to play.  Once the pressure was off the job hunt I really got to sit down to what I thought was going to be my first week off in almost 10 years.  (no-such-luck, I’ve got a job scheduled for tomorrow).

You can see that I paid attention to cable management, though it's still not as nice as I would have liked.

The Dell PV660F is supposed to be managed by Openmanage Array Manager.  Now the HCL on Array Manager is *SO* limited, that I had to go out of my way to find a pair of old QLogic 2200 cards, back-dated drivers (latest and greatest wouldn’t work) and it *STILL* only worked half the time.

So I do some web-surfing and I find this link:  ramblings of the village idiot » Blog Archive » powervault 660f management

This is where I make the connection – The Mylex controllers can be managed by the current incarnation of the Mylex management software.  (It’s noteworthy here that Mylex hasn’t been mylex in years.  From Mylex, to LSI, to Engenio, to Kevin Bacon)

So the solution to my problem is….

http://www.engenio.com/support/profibre_downloads.html

Yes, my desk is still a mess. :)

Now I’ve got everything up and running, the background init worked wonderfully, and I have 2 170G luns presented to a test VMWare server for playing with.  (It’s also noteworthy to point out that my first drive failed within 15 minutes of carving the lun, but thankfully I saw fit to configure two hotspares (knowing I was using used drives) and the hotspare invoked immediately and without issue.

So I’ll play with this for a week or so.  I have to build a new virtual-center host anyway (while not advised, a “virtual” virtual-center server is supported.) so I guess I’ll put it there.

Since all of this storage can also be made visible to the Windows host that runs Veritas, I’m smelling some play-time with VMWare Consolidate Backups – VCB for non VM People. 🙂

Aidan - watching videos on my new toy.

Anyway, I know that for most who read this this is boring stuff, but I think it’s cool.  So does my son, Aidan – pictured here watching YouTube videos of the Concorde.  In the grand sceme of things I’m a geek first – I only work to feed my obsession. (and feed the kids, that kind of comes in handy too)

8 comments

Skip to comment form

  1. Andrew Storrs

    Actually a “virtual” VirtualCenter is not only supported, for the most part these days it’s recommended. It requires a little bit of knowledge if the host running vCenter goes down (i.e. tell DRS to never VMotion your vCenter virtual machine or you’ll have to guess which ESX host to connect to – needle in a haystack in a large data center). By doing this it allows you to take advantage of VMware HA for increased availability and it allows you to take a snapshot prior to performing an upgrade – both are really handy.

    Have a look at this whitepaper for the benefits and caveats: http://www.vmware.com/pdf/vi3_vc_in_vm.pdf

    Duncan Epping also has some good tips here you should have a read through: http://www.yellow-bricks.com/2008/11/19/make-virtualcenter-highly-available-with-vmware-vi3/

    Dave Lawrence also blogged about it a few weeks back and covers some of the pros nicely: http://vmguy.com/wordpress/?p=67

    (Note: Both Duncan and Dave work for VMware as Senior Consultants/Systems Engineers)

  2. TimC

    @Andrew:

    So HA is occurring at the esx server level then? I was under the impression that HA was managed by the vCenter server itself, so if it were to go down, none of the VM’s would go anywhere, since there was nothing to manage the moves.

  3. Jesse

    That was always my objection to a virtual VC server. However the same limitation applies to a regular server so I’m not sure if there is any net loss of functionality.

    Without VirtualCenter, VMotion/DRS or VMotion/HA don’t work, right? Doesn’t that then mean that if the host that is housing the VirtualCenter VM dies the other hosts won’t / can’t take over for it?

    Unless as you say HA is set up at the host level, which given some of the configuration machinations that goes on when you configure it, wouldn’t surprise me.

    First and second hardware failures came out of the work yesterday.

    One drive failed – hot-spare invoked perfectly and even kept to the same bus so that the IO would remain balanced.

    Found a bad Emulex LP9802 HBA in one of the ESX hosts. Considering I got them for about $7.50 each, I’m not entirely surprised. 😉

  4. TimC

    LOL, gotta love ebay! Seeing your post motivated me to re-rack some of my own gear last night and try throwing on an expansion shelf only to find out I had incompatible hardware. BOO@that. Guess I’ll live with some 70GB LUNS for now.

  5. InsaneGeek

    ESX HA actually is a derivative of Legato AutoStart. It is installed onto the ESX servers themselves, and doesn’t require virtual center to be there to deal with a server failure.

  6. Jesse

    You are the first person, in or out of EMC, who has been able to really answer that question.

    Of course I’ve not been asking it as widely as I could have.

    I did my first “Storage VMotion” move last night. SWEET! Downloaded the GUI plugin for it because I’m getting lazy in my old age…

    Works like a champ.

  7. william bishop

    Jesse, if you have any questions (that was actually an easy one), you can go to the vmware forums (where most of us hang out). Ask a question, get an answer in 15 minutes!

    I’m glad your first dmotion went well, I sweated bullets the first few dozen or so, but now it’s SOP. I’ve only had two instances flake out on me, and out of the dozens upon dozens of times I’ve done this while in production (our systems can’t see downtime), I’m more than pleased with it.

    As to the very beginning, yes virtualizing VC itself is acceptable, if you are super paranoid, you can always set it up in a 2 node cluster of it’s very own (although it will cost you), or you can put it in a vm for portability, but only put the one vm on a host out of a pair of hosts, not managed by VC. Yes, you’ll have to start it manually if it dies on one host, but it’s not a big deal, as you can go days without VC if necessary.

  8. nc

    Hey, glad to hear that my blog post on the PV660 was helpful for you.. yeah, those things are beasts! I’ve still got one downstairs that I keep planning on eBay’ing – should get around to that one of these days. 😉

    -Nate

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>