Hyper-V 2 node cluster replacement

[NB9457]

Hmmm, interesting on that pricing, although I can't hit that price point on the configurator!

 

I am currently thinking, 2 hosts with 128GB RAM and around 8TB of usable storage after RAID, 5 should be fine, as the disadvantages of RAID 5 shouldn't be a problem with SSD's. Dual 10GBe, one for LAN, one for dedicated live migration/VM backup network.

 

Then a 3rd server with good old fashioned rust drives with enough capacity to hold the contents of both hosts to capacity using DPM, which we use already for our backups to home site and trickle backup to our offsite DPM server.

 

The simplicity is amazing lol but will fill all needs.

 

Plus the cost savings too, it does become harder to justify a SAN aha. Be interesting to know what you go with though, keep us posted!

 

Are there any brands you've been looking at in particular with regards to servers?

[CAWJames]

Plus the cost savings too, it does become harder to justify a SAN aha. Be interesting to know what you go with though, keep us posted!

 

Are there any brands you've been looking at in particular with regards to servers?

 

Mainly Lenovo and Dell,

 

The Dell 740 was coming out at £6k for a single Xeon Silver 4114 2.2G 10Core and 8 16GB RDIMMS and minimal storage.

 

The Lenovo x3650 with a single E5-2630v4 2.2G 10Core but 8 x 32GB came out at slightly less than the Dell.

 

I am sure shopping around I can get a similar server with dual procs for similar or less by visiting the supermicro builders dotted around! I would build my own, but management may frown if I can't provide 4hr response lol So builder with support offerings will be sought.

[NB9457]

Mainly Lenovo and Dell,

 

The Dell 740 was coming out at £6k for a single Xeon Silver 4114 2.2G 10Core and 8 16GB RDIMMS and minimal storage.

 

The Lenovo x3650 with a single E5-2630v4 2.2G 10Core but 8 x 32GB came out at slightly less than the Dell.

 

I am sure shopping around I can get a similar server with dual procs for similar or less by visiting the supermicro builders dotted around! I would build my own, but management may frown if I can't provide 4hr response lol So builder with support offerings will be sought.

Well, is 4 hour really necessary? We can live with next day, again, given the costs savings and the minimal impact of losing one host. You'd have to fail-over some services anyway regardless, so whether they stay running on a different host for 4 or 24 hours doesn't seem a huge problem. Back in the days of tin servers, I get the 4 hour things, but the beauty of VMs is, for me, in this environment with a good plan, you can just throw them here there and everywhere on any spare tin you have if you need to.

[CAWJames]

I was being a bit glib with the 4hr comment, but SMT and my manager would want some external support and expedited hardware replacement warranty to be in place for such a integral system to our way of working. It all has to hit our DR plan. As you say, next day replacement of HW would be fine as VM's can be migrated/retrieved from backup quite quickly given enough bandwidth and fast storage.

[Steve21]

Well, is 4 hour really necessary?

 

Pfft 4 how slow?!?

 

7Yr ProSupport and 2hr Mission Critical

 

Was playing around on the dell site earlier and found a new option 2 hour!

 

But yeah I know what you mean Especially if it's not a main server and it's all virtual etc

 

Steve

[CAWJames]

Pfft 4 how slow?!?

 

 

 

Was playing around on the dell site earlier and found a new option 2 hour!

 

But yeah I know what you mean Especially if it's not a main server and it's all virtual etc

 

Steve

 

7yr and 2hr! Wow that will be for the supportable life of the server, they won't go over 7 years, unless they have changed that for newer revisions of the hardware. Bet that isn't cheap!!

[CAWJames]

Right, an update on this, we have finally pulled the trigger on a solution.

 

We have opted for no SAN, and 3 servers.

 

Servers 1 and 2 are Dell 740XD with 16 1.8TB 10K rusty drives, and 2 240GB SSD;s, dual Xeon Silver 4114 and 128GB of RAM.

 

Server 3 is the same but with only 32GB of RAM.

 

Both have Quad 10GBe cards in and 3 yrs 4hr prosupport

 

Server 1 and 2 will be the Hyper-V hosts with a 10GBe crossover between them for SNLM

 

Server 3 will be the new DPM backup server, it has the same amount of drives as the others, as DPM uses block level, whereas the others will use RAID 10, so effectively 14.4TB useable on each of the hosts and 28.8TB in a JBOD for the backup server. There will also be a 10GBe crossover from each host to the backup server, to create a dedicated backup network for the system.

 

It has ended up costing more than we wanted, as we went for Dell with prosupport, but it is considerably less than a SAN based system.

 

Now to try and monetise an Equalogic PS6100x with a full compliment of 900GB drives for 21TB RAW, 2 Dell 720's with 128GB of RAM and dual E5620's and a Dell 2950 with dual E5420, 8GB of RAM and an MD1000 drive-shelf with around 14TB of storage.

[mukz]

Right, an update on this, we have finally pulled the trigger on a solution.

 

We have opted for no SAN, and 3 servers.

 

Servers 1 and 2 are Dell 740XD with 16 1.8TB 10K rusty drives, and 2 240GB SSD;s, dual Xeon Silver 4114 and 128GB of RAM.

 

Server 3 is the same but with only 32GB of RAM.

 

Both have Quad 10GBe cards in and 3 yrs 4hr prosupport

 

Server 1 and 2 will be the Hyper-V hosts with a 10GBe crossover between them for SNLM

 

Server 3 will be the new DPM backup server, it has the same amount of drives as the others, as DPM uses block level, whereas the others will use RAID 10, so effectively 14.4TB useable on each of the hosts and 28.8TB in a JBOD for the backup server. There will also be a 10GBe crossover from each host to the backup server, to create a dedicated backup network for the system.

 

It has ended up costing more than we wanted, as we went for Dell with prosupport, but it is considerably less than a SAN based system.

 

Now to try and monetise an Equalogic PS6100x with a full compliment of 900GB drives for 21TB RAW, 2 Dell 720's with 128GB of RAM and dual E5620's and a Dell 2950 with dual E5420, 8GB of RAM and an MD1000 drive-shelf with around 14TB of storage.

 

May i ask how much you paid for all this? Also how comes on 3yrs?

[colacao82]

I too would be interested in a ballpark figure

 

Also... what is SNLM?

Edited January 29, 2018 by colacao82

[CAWJames]

It all came in the low to mid 30's ex VAT, just over 10k a server.

 

Only reason for 3 years is we can make a judgment in 3 years if we want to up or down the level of support, never know what is going to happen!

 

SNLM is Share Nothing Live Migration, sort of like a cluster with no SAN, so you can still drain off VM's to the other host if needed, just not as quick. Our small servers will drain quickly, our file servers not so much! Will give it 2 10GBe pipes to play with and see how it goes! This still gives us the option of host maintenance with little to no downtime, which was one of the main reason for a SAN in the first place.

[PotNoodleTech]

Now to try and monetise an Equalogic PS6100x with a full compliment of 900GB drives for 21TB RAW, 2 Dell 720's with 128GB of RAM and dual E5620's and a Dell 2950 with dual E5420, 8GB of RAM and an MD1000 drive-shelf with around 14TB of storage.

 

Congrats on getting the system sorted. Those new dell servers will last 10 years.

 

Lol I just ordered a R720 to use as a Veeam Backup and Recovery server - one mans junk is another mans upgrade!!

 

The R720 are great servers and worth a few bob flog them on here to schools with less budget. The PE1950 and MD1000 are pretty worthless nowadays, why don't you use them as a training/test system? That's what I'd do!

[CAWJames]

That's how I sold the extra outlay to the powers that be, they will last longer, be quieter and use less electrickery.

 

The 720's have been rock solid as hyper-v hosts in a 2 node cluster, but alas they have hit their 7 year prosupport cut off, and as of May can't/won't be supported, and that makes people twitchy here!

 

The 2950 is 10 years old and as you say worthless really, it just has a crapload of storage that I can't find a use for, ie no live/production loads, and already have a test hyper-v system with oodles of storage and another 720 strapped to it.

 

#firstworldproblems really!!

[NB9457]

That's how I sold the extra outlay to the powers that be, they will last longer, be quieter and use less electrickery.

 

The 720's have been rock solid as hyper-v hosts in a 2 node cluster, but alas they have hit their 7 year prosupport cut off, and as of May can't/won't be supported, and that makes people twitchy here!

 

The 2950 is 10 years old and as you say worthless really, it just has a crapload of storage that I can't find a use for, ie no live/production loads, and already have a test hyper-v system with oodles of storage and another 720 strapped to it.

 

#firstworldproblems really!!

I can only dream!

[SHimmer45]

same here having that kind of kit "lying about" would open up all sorts of possibilities in testing and training

[sluggster66]

You could extend the warranty on those 720s with a third party support company for a few hundred pounds a year.

 

They will support the servers for at least another 5 years after Dell EOL

 

Simon

[CAWJames]

Yeah, I know, it's not so much the servers though, its the SAN that is a barnyard to get support on. Dell wanted something like £7k for a year of 4hr prosupport for it(taking it down to NBD was still £5k)!! Aftermarket was a bit better, but still low thousands. We took the business decision to refresh all the hardware and then try and recoup some pennies(hopefully many thousands of them) to offset later on when we have finished migrating. and they look so "pretty" with their new DELLEMC honeycomb fronts on. Did wish I had bought the bluetooth one touch setup system, but just wasn't worth it on 3 servers, 30 maybe but not 3!