Failover cluster with one SAN

[blu4]

Hi,

 

We are in the process of fishing out and studying possible virtualization options. We have been given a quote for two new servers and two new SANs. Is it a must to have two SANs? One for each node? I mean, it gives you the ultimate redundancy but the SAN has got redundancy of its own like RAID 6 in two different arrays. Any suggestions? Getting one SAN less will help our budget a LOT. I have attatched a picture of what I have in mind.

 

 

Top switch is for client communication with the nodes, middle one is for heartbeats and the bottom one is for node storage access. The SAN at the bottom will be a CSV.

 

Thanks

[Achandler]

Yes it is perfectly possible.

 

If you get a SAN with 2 controllers then you have even better redundancy.

 

Also have connections from the SAN and servers going through 2 switches, if your using iSCSI then this is easy. Then you remove the switches as single points of failover as well.

[blu4]

Yes it is perfectly possible.

 

If you get a SAN with 2 controllers then you have even better redundancy.

 

Also have connections from the SAN and servers going through 2 switches, if your using iSCSI then this is easy. Then you remove the switches as single points of failover as well.

 

Thanks for the answer. I would like to stick to one SAN since it saves about 7k. In addition, we have got an 8TB NAS- would it also be possible to leverage that as an incremental backup for the SAN?

[Achandler]

For the SAN directly, I don't think so. But depending on your backup routine you could use it as a disk for DPM to use and allow you to perform bare metal recovery on your servers, 8TB would probably have backups going back upto 2 weeks (depending on data).

 

If you do go for one SAN, like I said check if it has 2 controllers, it will make the 1 SAN more, but obviously not as much as 2 SANs.

[sonofsanta]

We run a HyperV failover cluster on a single SAN with dual controllers, interconnected through two switches for complete redundancy. I don't know why you'd go two SANs instead, to be honest, certainly not on a school budget (and there are people here who question spending on one).

[blu4]

We run a HyperV failover cluster on a single SAN with dual controllers, interconnected through two switches for complete redundancy. I don't know why you'd go two SANs instead, to be honest, certainly not on a school budget (and there are people here who question spending on one).

 

When I heard the quote for the hardware, I was shocked. Thanks for all the input guys. Little but of research and help from you just saved us about 9k. Thanks!

[Achandler]

I worked in both, one SAN two controllers and 2 SANs.

 

2 SANs has the advantage or you can have effectively 2 server stacks running in different locations, so should there be a catastrophic failure in one area of the school, (fire, power cut, water) then effectively you coudl run off the 2nd location. In reality though most school networks would have a single central location where most links go back too, so although you may be able to continue operating, it would be on a vastly reduced capacity and for vital services only. The two SAN solution was already in place when I arrived.

 

2 SANs also really does need 2 locations, both with aircon and direct fibre links between them to make it even worth considering, which isn't practical in a lot of places.

[zag]

Try not to discount local storage so quickly.

 

For 7K you could get 6 brand new servers with local storage offering far better redundancy.

 

For 14K you could have mirrored server rooms!!

 

SANs are not always the best option for small organizations like schools with less than 50 servers

[tmcd35]

Try not to discount local storage so quickly.

 

For 7K you could get 6 brand new servers with local storage offering far better redundancy.

 

For 14K you could have mirrored server rooms!!

 

SANs are not always the best option for small organizations like schools with less than 50 servers

 

^This,

 

If I was to start 100% from scratch now I thin I'd look at local storage. A combination Hyper-V Replica and DFS-R shares seem to offer all the redundancy and failover support that I need in this environment.

 

As per other threads, we are extending our network to two storage servers and using Hyper-V Replica and DFS-R to replicate between them. Our set up is essentially two Windows based NAS boxes in a SAN-like configuration (a separate "storage network" for SMB traffic between Virtual Hosts and the NAS). I've been getting quotes between £7k and £11k for a Windows Server with 8-10Tb of storage.

[blu4]

^This,

 

If I was to start 100% from scratch now I thin I'd look at local storage. A combination Hyper-V Replica and DFS-R shares seem to offer all the redundancy and failover support that I need in this environment.

 

As per other threads, we are extending our network to two storage servers and using Hyper-V Replica and DFS-R to replicate between them. Our set up is essentially two Windows based NAS boxes in a SAN-like configuration (a separate "storage network" for SMB traffic between Virtual Hosts and the NAS). I've been getting quotes between £7k and £11k for a Windows Server with 8-10Tb of storage.

 

Am I correct in understanding that you just populate the servers with drives and replicate across the network with DFSR?

[tmcd35]

Am I correct in understanding that you just populate the servers with drives and replicate across the network with DFSR?

 

Hyper-V doesn't work with DFS-R namespaces so you have to use Hyper-V Replica instead on standard SMB shares (very, very easy to set up). 2012R2 allows replication between three Hyper-V servers, 2012 is just 2.

 

The DFS-R shares are for our home drives, and other user shared folders. Users connect to the storage servers through on bonded network link, the VHD's are connected through a separate link to segregate traffic and reduce load/bottlenecks.

 

All in all work quiet well for us and with SMBv3 provide the speed we need. Just another alternative to iSCSI and CSV.

[blu4]

Hyper-V doesn't work with DFS-R namespaces so you have to use Hyper-V Replica instead on standard SMB shares (very, very easy to set up). 2012R2 allows replication between three Hyper-V servers, 2012 is just 2.

 

The DFS-R shares are for our home drives, and other user shared folders. Users connect to the storage servers through on bonded network link, the VHD's are connected through a separate link to segregate traffic and reduce load/bottlenecks.

 

All in all work quiet well for us and with SMBv3 provide the speed we need. Just another alternative to iSCSI and CSV.

 

Have you got any website or tutorial resource explaining that further?

[tmcd35]

Here's a rough diagram of our current network topology:

 

 

We actually only have one storage server at the moment, set up with SMB shares only. We are looking at purchasing the second now and reconfiguring around DFS-R for user shares and Replica for VHDs.

 

You can, under this model - as @zag suggests, forget the storage servers completely and have the storage built in to the host servers - doing away with the need for a storage network. Since we're expending and not starting from scratch, we're going down the second SMB storage server route. Looked at iSCSI but couldn't make it work for us with our existing storage server which we don't really want to get rid off.

 

Here's MS's guide to running Hyper-V over SMB shares: Deploy Hyper-V over SMB

A good technet article on Replica: Hyper-V Replica Explained, Windows Server 2012 Style - Yung Chou on Hybrid Cloud - Site Home - TechNet Blogs

And DFS Replication: DFS Namespaces and DFS Replication Overview

 

I'm looking at no hardware RAID on the storage boxes and using Windows 2012R2 Storage Spaces to create the partitions needed to store each shared folder (VHD, Replica, Home, Staff, and Students): Step-by-step for Storage Spaces Tiering in Windows Server 2012 R2 - Jose Barreto's Blog - Site Home - TechNet Blogs

[PotNoodleTech]

The SAN is much more likely to fail compared to either a switch or a controller card so it makes no sense to me whatsoever, to double up on switches and controller cads and then just use one san. Doesn't matter how many switches you throw at is, that is not a redundant system.

 

That being said, dual redundant sans, switches, controllers, server rooms, air on units, UPSs, network managers, etc are way beyond the budget of most schools, and I would think very carefully before spuning £50k up the wall as I suggest it's overkill given 1) the low likelihood of a catastrophic failure of the equipment 2) the superfast super efficient data restoration now offered by Software like Veeam Backup and Recovery which to a certain degree negates the necessity for redundancy (other than on disks).

 

My refresh in 2015 is quite likely to have local storage but use Vmware VSAN to leverage the redundancy to the next level all backed up via Veeam. Short of a nuke, that would offer me 99.9% uptime for under £15k probably.

[kevin_lane]

For us we have 3 host and a san we most probably have in the region of about 15 servers some that are created and some p2v with dpm and veeam I suppose also you need to work out how many servers your going to be needing we have a 24 TB san which will see us through for a long time but you also need to work out ur growth because if you can do something without spending an extra 10 grand ur bursar will love you also have you looked into 2012 r2 as now they offer a cluster in a box solution which could save you lots you just need to make sure your front end servers are a good spec I think also if you know your environment and what you need then you should go with what feels right and there is nothing worst than putting in a system where you have cut sort because you wasnt allowed it or dont have the budget obviously there is limits because a schools budget dont strech alot but also you have to work to a solution that you know that if you have a powercut your host(s) is going cope and the services are going to be running it took me best part of 2 years to make sure that the equipment and specification of what I wanted to achieve was bang on and if you want to go down the virtualised route you have to do your home work because you need a solid foundation other wise its going to come down like a tonne of bricks if you get it wrong not to scare you though lol

[blu4]

Here's a rough diagram of our current network topology:

 

[ATTACH=CONFIG]27175[/ATTACH]

 

We actually only have one storage server at the moment, set up with SMB shares only. We are looking at purchasing the second now and reconfiguring around DFS-R for user shares and Replica for VHDs.

 

You can, under this model - as @zag suggests, forget the storage servers completely and have the storage built in to the host servers - doing away with the need for a storage network. Since we're expending and not starting from scratch, we're going down the second SMB storage server route. Looked at iSCSI but couldn't make it work for us with our existing storage server which we don't really want to get rid off.

 

Here's MS's guide to running Hyper-V over SMB shares: Deploy Hyper-V over SMB

A good technet article on Replica: Hyper-V Replica Explained, Windows Server 2012 Style - Yung Chou on Hybrid Cloud - Site Home - TechNet Blogs

And DFS Replication: DFS Namespaces and DFS Replication Overview

 

I'm looking at no hardware RAID on the storage boxes and using Windows 2012R2 Storage Spaces to create the partitions needed to store each shared folder (VHD, Replica, Home, Staff, and Students): Step-by-step for Storage Spaces Tiering in Windows Server 2012 R2 - Jose Barreto's Blog - Site Home - TechNet Blogs

 

How many users/workstations do you currently have?

[tmcd35]

How many users/workstations do you currently have?

 

So, current set up - 1 storage box, 4Tb (450Gb 15k SAS drives in RAID-50), Windows Server 2008R2 (SMBv2), 3 virtual machine host servers running 2012. About 30 virtual servers in total (2 terminal servers, 2 SIMS servers, Exchange, several SQL and Internet servers, etc). About 250-300 client machines. 750 Pupils. 100 Teachers.

 

The (current) plan is to move to 2 storage boxes of 8Tb each (combo of SSD, 900Gb 10k SAS, 450Gb 15k SAS) and Server 2012R2 for the storage. VM's will be manually load balanced between the two storage boxes with Hyper-V Replica copying the VM's to the non-live storage box for the respective VMs. DSF-R will allow us to load balance user file shares between the two storage boxes. The only great unknown at this stage is the bandwidth requirements for both Replica and DFS-R.

 

 

I will say I'm not trying to promote this method as a better solution to iSCSI. It's just a good alternative that was the right fit for us. In many ways, if you can afford it, iSCSI and CSV may have some advantages. I suppose it depends on the size of the organisation. We're a small to medium sized rural high school, to dropping £50k on the SAN to end all SANs was never an option for us.

[blu4]

So, current set up - 1 storage box, 4Tb (450Gb 15k SAS drives in RAID-50), Windows Server 2008R2 (SMBv2), 3 virtual machine host servers running 2012. About 30 virtual servers in total (2 terminal servers, 2 SIMS servers, Exchange, several SQL and Internet servers, etc). About 250-300 client machines. 750 Pupils. 100 Teachers.

 

The (current) plan is to move to 2 storage boxes of 8Tb each (combo of SSD, 900Gb 10k SAS, 450Gb 15k SAS) and Server 2012R2 for the storage. VM's will be manually load balanced between the two storage boxes with Hyper-V Replica copying the VM's to the non-live storage box for the respective VMs. DSF-R will allow us to load balance user file shares between the two storage boxes. The only great unknown at this stage is the bandwidth requirements for both Replica and DFS-R.

 

 

I will say I'm not trying to promote this method as a better solution to iSCSI. It's just a good alternative that was the right fit for us. In many ways, if you can afford it, iSCSI and CSV may have some advantages. I suppose it depends on the size of the organisation. We're a small to medium sized rural high school, to dropping £50k on the SAN to end all SANs was never an option for us.

 

Well, if the governors will listen, then I will go in a similar path to you. But hey, I doubt they will listen to a 21 year old. :<

[kevin_lane]

of course they will look all you need is a plan you just need to make sure you have covered all your basis you also have a good plan of how to put it all together also have you decided on what platform you want to do this all on ? Dell or HP ?

[blu4]

of course they will look all you need is a plan you just need to make sure you have covered all your basis you also have a good plan of how to put it all together also have you decided on what platform you want to do this all on ? Dell or HP ?

 

We got a quote from a supplier for a beefy HP server however, I am still doing research and would like opinions for virtual host platforms.

[tmcd35]

You'll almost certainly need quotes from at least 3 suppliers all to the exact same specification. This last round went out to 5 suppliers - Insight, Misco, Stone, @VeryPC, and Novatech. With some very interesting results. Our last 3 servers (the virtual hosts) and our original storage box all came from Novatech. I understand the brand attraction to HP and Dell but you shouldn't overlook what bespoke suppliers like Novatech and @VeryPC can offer.

 

As for your governors, are you employed as the schools IT expert? If so then they should listen - are they IT qualified and know the details of network storage better than you? Just write them a clear concise report that details all the different options, the merits and pitfalls or each, and your recommended option. Keep the language to something laymen could digest and technical detail to a minimum and you'd be surprised how well they'll listen (or at least should listen).

[nicholab]

Don't forget a solution based on a Dell VRTX.

[kevin_lane]

well depends on what you feel comfortable with do you want to use hyper-v, we use hyper-v and works like a dream however you also have vsphere(VMware) I think you have to figure out which is 1 going to be easier to implement 2 how easy it is to configure 3 how much does cost ? 4 how are your backups/fail over going to work. you also need to make sure that when you are looking at putting a virtual host you need to make a guess of your environment I.e what is your current infrastructure are you on 100mbp back bone or are you on 1Gbps and are you going to have enough bandwidth to cope with the amount of data, for us (not saying you should do this) but I'm in the process of replacing all of out 1Gbps switches over to 10Gbps but that is because I want future proof us, and plus we are getting bigger but all the end points will only be 1Gbps, and in the process of setting up a new 4th virtual server so I can export my hyper-v into the new 2012r2 and migrate our cluster (looong process esp when you don't have funds and time) but the hardware for your hosts is it going to be enough and how many vms are you planning on putting on there, also if you have any physical servers that you want to v2p have a think and see if they actually need 3 processors or do they need that extra 2 gigs of ram but also you need to plan a disaster plan and think if 1 of my host go down can the other host take all of the vm most prob said this before earlier up but its good think about it

[zag]

I'd ask the question why bother with replication and load balancing?

 

With virtualization and Veeam instant backups it takes a matter of minutes to restore a VM to any storage layer.

 

There really is no need for it these days imo unless you are running some kind of ultra critical infrastructure.

 

I do understand that if you are putting all your eggs in one basket (i.e. a SAN) then this does become more important, but I still question the need for replicated volumes due to the cost.

 

The only performance bottleneck I've ever seen on Hyper-V is RAM

[blu4]

I'd ask the question why bother with replication and load balancing?

 

With virtualization and Veeam instant backups it takes a matter of minutes to restore a VM to any storage layer.

 

There really is no need for it these days imo unless you are running some kind of ultra critical infrastructure.

 

I do understand that if you are putting all your eggs in one basket (i.e. a SAN) then this does become more important, but I still question the need for replicated volumes due to the cost.

 

The only performance bottleneck I've ever seen on Hyper-V is RAM

 

Are you using Hyperv and veeam? I just had a look at it and looks promising, however I'd like first hand opinion about it.