Jump to content

[help] virtual Clustering?


fizban2

Recommended Posts

Ok,

learning more about clustering but want to try and build my own cluster, anyway to build one with getting mutliple machines for it? was thinking of using virtual server 2005 and some VMs with 2k3EE and try getting them into a cluster. any thoughts on this?

Link to comment
Share on other sites


First of all, Virtual Server 2005 R2 isn't slow unless you're running it on slow hardware - it's a server product, and it needs a server-class machine to run on (if you wouldn't run SQL on it, you shouldn't run VS 2005 on it!). VS 2005 works very well when it's run on good hardware, and each VM does not need the memory footprint that a VMWare VM needs to run at similar speeds once booted.

You can use the following articles to help you cluster 2 machines in VS 2005 (although you can only cluster 2 virtually, due to limitations in the virtual SCSI adapter used in VS):

http://www.microsoft.com/technet/prodtechn...VM_cluster.mspx

http://www.microsoft.com/technet/prodtechn...oy/cvs2005.mspx

It's quite easy to do using the instructions in the articles above.

Link to comment
Share on other sites

You are entitled to your opinion, and I would always suggest people use products that work best for them in their environment. However, just saying a product sucks without pointing out specific reasons isn't very good advice.

Edited by cluberti
Link to comment
Share on other sites

thanks all for the advice and links,

@Memnoch: awesome guides, just what i need,

@jondercik: VS2005 is slow on desktop HW but trying to get a trial of ESX is harder to do, ESX is a good server product but doesn;t run any faster on a desktop rig then VS2k5 will.

@cluberti: thanks for the specifics on the memory thing, didn't know that. I will assume you have built clusters before, would it be possible to PM you if i run into issues? also the hardware i would be testing this on would be a p4 2.4 w/HT, 2 gigs ram, 2 x 120 gig SATA drives. i know it is gonna be kinda slow to run, i know i can get it going :)

Link to comment
Share on other sites

  • 3 weeks later...

I have a question along the same lines as this thread.

I've got an older Dell PowerEdge 6400 that has Quad PIII 900/2MB CPUs, 8GB RAM and 8 x 36GB U160 10K RPM hard drives. Will this be hardware enough to run two virtual servers for a virtual SQL cluster...strictly for testing and development? I'm thinking it will be given the limited number of users that would be hitting it but I thought I'd ask people who have more experience with virtual servers than I do. I've done several clusters, just none in a virtual server environment.

Link to comment
Share on other sites

The only time it would be under what we consider "heavy" load is if the main cluster had a serious catastrophy...something that would prompt me to rebuild the entire thing from scratch. Something we consider "heavy" for our setup is about 25, maybe 35, concurrent users. Our setup is strictly for failover instead of performance. The hardware is adequate enough that we have three instances all running on one box while the other does nothing unless the first one fails (which has never happened but you never know). I'm trying to talk them into purchasing the additional SQL license so we can do an Active/Active setup. Then I'd move the two lesser used instances to the 2nd node while the 1st node runs the larger instance.

I'm actually thinking about using a virtual server setup as a staging point to reinstall the physical servers. We did an in-place rolling upgrade from Windows 2000 Advanced Server to Windows Server 2003, Enterprise Edition...and I hate upgrade installs. We're also looking at migrating from SQL 2000 to SQL 2005 in the near future so I'd like to have it all on a clean setup. :)

Edited by nmX.Memnoch
Link to comment
Share on other sites

...so we can do an Active/Active setup.

Then it's not really a cluster at that point, it's more of a robust "load balancer". Clustering is supposed to provide failover and uptime, not extra performance :).

If you're going to do active/active, then you should have two passive nodes to fail to. Just a pet peeve of mine, ignore me, I go away eventually ;).

Link to comment
Share on other sites

Lol, yeah memnoch, just tell them you need another 3 servers... they won't care ;)

@cluberti, yeah not having any passive nodes does make the cluster nothing more then a server on steriods, but budgets usually dictate the system used vs what most people actually need. and don't go away, we still have to pick your brain more :P

Link to comment
Share on other sites

...so we can do an Active/Active setup.

Then it's not really a cluster at that point, it's more of a robust "load balancer". Clustering is supposed to provide failover and uptime, not extra performance :).

It can still do failover as long as you make sure that both nodes are powerful enough to run all of the instances should one of them fail. :)
If you're going to do active/active, then you should have two passive nodes to fail to. Just a pet peeve of mine, ignore me, I go away eventually ;).
Doing it that way would, in effect, give me two Active/Passive clusters instead of one Active/Active cluster. :D

Honestly though, I wish I could add more nodes but they wouldn't spring for a fiber channel drive setup when we initially purchased 2 years ago. My budget was $50K to get it up and running...I think I did pretty good considering the budget. The initial system was two Dell PowerEdge 6600's w/ Quad Xeon MP 1.5GHz/1MB cache and 4GB RAM. With those we purchased a PowerVault 220S for the shared storage and loaded it with drives. We later upgraded the servers to 8GB RAM and added another PowerVault 220S (each database has two drive sets...one for data, one for logs...plus the Quorum drive and another drive for MSDTC). We have plans to upgrade the CPUs to Xeon MP 3.0GHz/4MB cache and 16GB RAM later this year...providing the budget is there. When we replace the systems in a few years I'll talk them into fiber channel to go with the new servers.

Oh, and my other post was supposed to say "migrating from SQL 2000 to SQL 2005"...not "from SQL 2000 and SQL 2005". Doh...

Edited by nmX.Memnoch
Link to comment
Share on other sites

I know budgets usually dictate IT actions, but I still need to throw that in there. And yes, it is correct that active/active works if both boxes can handle the load of both servers on one particular node, but there's always the problem of creep over time - more databases, more users, etc, and over time you run into a cluster that can't quite keep up running both node's load on one node. I've seen it hundreds of times, and always having one passive node for one active node solves that problem (yes, it's more expensive, but you are using a cluster to guarantee uptime, ultimately - not improve performance).

Doing it that way would, in effect, give me two Active/Passive clusters instead of one Active/Active cluster

Not really. Since any active node can fail to any passive node, there's no "this node can only fail to that node" issue. The bigger the job, the more nodes you should have - but always one passive for every one active node (up to 4 active / 4 passive, obviously). Clusters are actually very good for patching scenarios - you fail to a passive node, patch the active, reboot, then fail back and patch the passive. As close to 100% uptime as one can get without not patching every month or so :).

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...