Would you virtualise?

[twin--turbo]

Both SQL has an upper limit, but we still use dynamic on the server (approx 6-8GB more than the SQL set) so it can fluctuate on other bits like backups/groupcall/parentpay processes blah blah

 

Steve

 

SQL will just use all the available memory unless you limit the instance, but it's not a problem it will release memory as the OS needs it for other operations.

 

One of our SQL servers has 32GB of ram with 6 DB instances each pegged at 5GB leaving 2GB for the OS.

 

 

one of the big + of virtualisation for us is we can spin up a test server in half an hour and burn it if we don't need it.. No Pleading to management for servers, networking..

 

Also when we came to upgrade our 70+ Servers in the summer we live migrated them from Old Dell R810's to R730's only hardware build was the 3 ESX machines. everything else was just migrated to new hardware with 5yrs support.

 

TT

[nicholab]

How much is the schools licence of sccm virtual machine manager? As I heard people suggesting vmware essentials us similar price? What share storage are people using there is scale out file server, datacore or Hp VSA or hardware (nimble)? Edited January 27, 2016 by nicholab

[localzuk]

How much is the schools licence of sccm virtual machine manager? As I heard people suggesting vmware essentials us similar price? What share storage are people using there is scale out file server, datacore or Hp VSA or hardware (nimble)?

 

SCVMM isn't expensive (don't know the actual cost), but it is also unnecessary. I just use the Failover Cluster Manager. SCVMM is more aimed at large deployments or VDI.

 

For storage, I just use SMB shares. Nothing fancy as yet. Next time I upgrade, I will be looking at greater resilience though.

[twin--turbo]

For storage, I just use SMB shares. Nothing fancy as yet. Next time I upgrade, I will be looking at greater resilience though.

 

We are NFS, faster and simpler.

[twin--turbo]

vmware essentials us similar price??

 

I think we were quote £750 for Essentials PLUS. that's for up to 3 hosts.

 

TT

[localzuk]

We are NFS, faster and simpler.

 

I dunno about simpler. NFS is pretty much guaranteed not to be a Windows box. Down here, getting people with skills other than Windows is difficult, so we have been attempting to keep things as simple as possible by using Windows servers for as much as possible. There's nothing much simpler than "Create Share"...

[twin--turbo]

Sorry we share backend storage as NFS to the VMware and Oracle Clusters from our SAN.

 

TT

[localzuk]

Sorry we share backend storage as NFS to the VMware and Oracle Clusters from our SAN.

 

TT

 

Ah, when we set this up 4 years ago, the SANs on the market were expensive and relatively inflexible (ie. if you wanted SSD storage, and SAS storage, you had to buy a tray of each, when we only needed half a tray of each). The new generation of storage devices seems to be better now though.

[twin--turbo]

So much choice now.

 

ours runs 3x24 disk shelves, 4 disks are SSD for cache. 30TB ish useable.. Combined with upgrading to 10Gb it has cut many processing jobs to a quarter of their old time

 

TT

[Steve21]

SQL will just use all the available memory unless you limit the instance, but it's not a problem it will release memory as the OS needs it for other operations.

 

One of our SQL servers has 32GB of ram with 6 DB instances each pegged at 5GB leaving 2GB for the OS.

 

Just to confirm when I said it has a limit I meant we set a limit on the SQL, then leave the other bit dynamic for other processes/services running on the same box

 

Steve

[andyturpie]

We have a fully vitrualized CC4 estate with two HP Proliant G8 hosts containing 192gb RAM in each box. A 16tb HP San takes care of everything storage wise with a 16tb RM disk box and LTO5 added for D2D2T.

 

We have virtualized all servers including SIMS set to max 16gb RAM and we have had no issues with this set up. We have a current uptime of 99.6% availability in the last 3 years. Would have been 100% uptime but a faulty GBIC took down a portion of the site for a day and the day after a pupil caused a data storm accidentally which took an hour to locate (both events happened two weeks ago lol).

[zag]

Yep same as everyone else really, virtualized now for years. Can't imagine going back to physical servers.

 

Sims admin server has 12gb of static RAM. All servers local SSD storage. Couple of spare hosts just in case sitting there doing nothing.

[twin--turbo]

I am waiting for my 5 new R730's to arrive some time today..... 256GB and dual 18 core.... not got round to shiffting stuff in the testing rack to get them in.

[hallb15]

I am waiting for my 5 new R730's to arrive some time today..... 256GB and dual 18 core.... not got round to shiffting stuff in the testing rack to get them in.

That's some serious kit - where do you work?!

[zag]

I am waiting for my 5 new R730's to arrive some time today..... 256GB and dual 18 core.... not got round to shiffting stuff in the testing rack to get them in.

 

Be careful putting all your eggs in one basket with large servers like that.

 

We have around 8 hosts with 3 VM's on each. That way if you have hardware failures, its actually pretty easy to restore services quickly.

[cpjitservices]

Everything virtual... Any servers you have get them virtualized its really easy. You can use Microsoft Disk2VHD tool if you go the Hyper-V Route. We use Hyper-V across our virtual cluster. We have 5 Dell R610's running Server 2012 DC with Hyper-V all replicated on and off site.

[Gibson335]

Be careful putting all your eggs in one basket with large servers like that.

 

We have around 8 hosts with 3 VM's on each. That way if you have hardware failures, its actually pretty easy to restore services quickly.

 

Don't really understand this. We have 2 physical servers, with around 15 VMs on each...but each physical has enough capacity so that in the event of physical server failure, the VMs can be shifted across to the other one and run quite comfortably.

[cpjitservices]

Don't really understand this. We have 2 physical servers, with around 15 VMs on each...but each physical has enough capacity so that in the event of physical server failure, the VMs can be shifted across to the other one and run quite comfortably.

 

Have you got replication setup between the two hosts ?

[localzuk]

Don't really understand this. We have 2 physical servers, with around 15 VMs on each...but each physical has enough capacity so that in the event of physical server failure, the VMs can be shifted across to the other one and run quite comfortably.

 

Indeed, that's how we run it too. There's zero need to have that many physical servers - it defeats the point of virtualising!

 

We run 26 VMs across 2 nodes. If I need to take one node down, or one crashes, the other can run the remainder of the VMs just fine. Never seen the CPU go above 50% averaged over a minute, even when everything is running on one node!

 

Only thing that has ever limited us is the RAM - and I am expanding that shortly to handle a new requirement we have of our cluster.

[cpjitservices]

Indeed, that's how we run it too. There's zero need to have that many physical servers - it defeats the point of virtualising!

 

We run 26 VMs across 2 nodes. If I need to take one node down, or one crashes, the other can run the remainder of the VMs just fine. Never seen the CPU go above 50% averaged over a minute, even when everything is running on one node!

 

Only thing that has ever limited us is the RAM - and I am expanding that shortly to handle a new requirement we have of our cluster.

 

How does that work then ? If you have replication setup does it automatically fail over ?

[localzuk]

How does that work then ? If you have replication setup does it automatically fail over ?

 

The VM storage is a separate device, and both Hyper-V nodes share that storage as a cluster. So, the storage doesn't need to be replicated at that stage. If one of them fails, the other knows this and grabs the rest of the VMs, starting them back up. Longest downtime is about 5 mins as the servers boot back up.

[cpjitservices]

Thats how we have it setup all our Hyper-V servers share a SAN basically where VM's etc are stored. How did u set it up so that the other server knows that the VM's have gone down on the other ?

 

I have heartbeat enabled is this all that needs to be done ?

Edited January 27, 2016 by cpjitservices

[truebluesteve]

We have 3 Dell 730's each with 128GB attached to an HP and EMC SAN. There are about 16 VM's across the hosts including SIMS, SQL Server and 2 x DC's. We don't have an Exchange server and the only box not virtualised is our legacy backup server which connects via SCSI to a tape library.

 

We've had this setup for 4 years or so now and its been extremely reliable so I'd say go for the whole lot!

[Gibson335]

Have you got replication setup between the two hosts ?

 

No need - even if the master fails you can use the command line on the other to seize control, and once you have the console you can reassign the VMs.

[localzuk]

Thats how we have it setup all our Hyper-V servers share a SAN basically where VM's etc are stored. How did u set it up so that the other server knows that the VM's have gone down on the other ?

 

I have heartbeat enabled is this all that needs to be done ?

 

I have a Hyper-V cluster set up via Failover Cluster Manager, with both the nodes added. The cluster then has a witness share set up on shared storage also (this is to ensure there's a non-even number of servers to determine quorum for the cluster). All VMs are set up as "Roles" within the Failover Cluster, some with preferred servers specified (eg. so as to spread our RDS servers between the 2 physical servers).