Hitting maximum number of LUNs

[NuggetGTR]

Hi all,

Just thought i would bounce an idea or 2 off some people here.

Basically looks like I will be hitting the 256 LUN limit for ESX, The issue is to do with P2Ving all the MSCS boxes.

I have a cluster which looks like it will be running out of space for LUNS, this cluster, one of many i manage is massive 32 high end hosts so we were already hitting 190 LUNS for the VM datastores.

early this week I was told they want to move all their MSCS clusters(currently consolidating as much as we can) The issue here is that each of the 50 clusters have 4 to 10 LUNS each(disk resources for MSCS)

What would be the best way to tackle this?

I was currently thinking along the lines of creating a new cluster of ESX hosts which will only hold the MSCS servers so it would give me full 256 LUNs to work with.

unfortunatly our data stores are at 500GB so ive been pushing to get 1TB LUN's assigned for datastores but this is a working progress, while this will help it doesnt solve the issue just delay it a little.

There isnt anyway to assign the LUNS directly to the Guest and not count towards the ESX LUN limit is there? Im using RDM's for the MSCS disk resources and as far as im aware these count towards the ESX LUN limit correct?

Cheers all

[idle-jam]

i would advise creating cluster of X number of Hosts and that each of these hosts would see that LUN only. but im curious how did you everk reach so many LUNS? are you using RDM instead of VMDK?

[NuggetGTR]

Yeah thats the line i was going along.

Im using RDM's for the MSCS resource LUNs as i mentioned. I not sure if this counts towards the ESX LUN number? if it doesnt then its all good but i assumed it would but havent found anything to say it doesnt.

With the huge hosts and being able to have 8 -12 cores per cpu and having 4 cpus, consolidation is getting massive and with 32 hosts in a cluster it would not take too long to hit the 256 limit in general... obviously make new clusters etc are ways around it. and we were fine untill i was told we would be virtualizing MSCS clusters.  but i really just want to know if RDM's count towards the ESX limit of 256 LUNs.

[AureusStone]

Yeah with MSCS you need to use RDMs.

I don't think there is any real workaround.  You just have to create smaller clusters unfortunately.

Edit: Yes RDMs count towards the maximum

[bulletprooffool]

I agree with the comments about LUN allocation, however I would very carefully consider whether this cluster is in fact a good candidate for virtualisation.

Remember, VMware is not always the solution.

I know this does not really address the original question, but I think you should maybe look at your Architectural design first.

You have further option:

Stay hardware only?

VMware FT?

Split the cluster into smaller clusters (as you are VMing this, your hardware costs are not affected, only your MS licensing costs)

and so on . .

good luck

[NuggetGTR]

Yeah i know virtualisation is not always the solution, but i dont think it came across right, I look after many large ESX clusters each cluster represents the different levels/network in the oginisation as its very(extremely) large so uat, test, dev, production etc.

I have 25 MSCS clusters coming over all in the same level about 50 - 60 servers(only the first lot to, there is more to come) they are not big boxes and virtualisation fits them well. its just because each MSCS cluster need 4 to 10 LUNS(RDMs) my ESX LUN limit is going to max out quickly and just wanted to bounce some sugestions off people here.

Cool so the basic consensus is that creating another cluster for the MSCS in that level or splitting up the current large cluster. thats what i thought but never know. lucky we have spare hosts to throw at this.

Cheers

[AlbertWT]

How about creating the future cluster with eagerzeroedthick VMDK as opposed to RDM ?

I read from this article and it is possible: http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=103582...