Virtualising SIMS (and other servers)

[mikkydoos]

I don't mind being shot down, the point is if you don't ask you don't know. As I say, Hyper V solves problems but then creates problems. Be interesting to actually talk to someone who's restored a Hyper V configuration (the actual host and more), but it's interesting that most (it appears) operate a clustered approach, rather than a standalone configuration - it makes sense but then the initial costs are probably high...

 

Again it's one of those situations where you question is buying 2 'super' clustered servers the way to go, or you could possibly buy 5/6 more reasonably priced servers?

 

 

Ure in Brum. Pop over and have a look in the new year if u want.

We're a primary but same setup wherever you are. Just a question of scale.

[Norphy]

Are you clustering with VMWare there @Norphy. Whats the product that looks after that failover ?

 

Yes, we're using VMware at the moment. We are licensed for vSphere. The part that looks after automated failover is DRS.

 

We are running vSphere for the moment but we are switching to Hyper-V, probably in the summer, as the licensing is much cheaper and there isn't much that VMware gets us which Hyper-V and System Center doesn't. The same principal, however, applies. I plan to set up failover clusters with Hyper-V so that VMs are automatically failed over to other hosts should there be a failure.

 

That's interesting, as of course Hyper V is basically its own OS, whereas VMWare is tiny. A totally different design.

 

With Hyper-V you solve one problem, but then create half a dozen, which is why I ask the question.

 

Hyper-V is a type 1 hypervisor, just like VMware. If you've got any sense, you'd run it on a Core instance of Windows and let System Center do the actual management of the product.

 

There no point to backing up your virtual hosts, at least in a failover configuration. Should a host fail, you buy another, install your hypervisor, add it to your cluster and get on with your day.

[Roberto]

That's interesting, as of course Hyper V is basically its own OS, whereas VMWare is tiny. A totally different design.

 

I'm honestly unsure what you're trying to say here. VMWare ESXi, and HyperV Server are both operating systems. They both provide a bootable environment which runs a "bare metal" hypervisor. How are they a totally different design?

[3s-gtech]

I virtualised our old physical SIMS server onto an ESXi host, where it was the only machine on the host. 30GB of RAM and 8x15k SAS in RAID 10. Worked fine, despite my first experience of doing this to an underpowered host not working out well. Backup was via usual SQL backups. Since then, moved it to another co-hosting setup with some other VMs on a more powerful host, running on SSDs. Much less RAM now. Working really well. Backups via SQL and now via GhettoVCB too for the full VM.

[CAWJames]

DPM is our go to for full backup of VM's and DR. Clustered at base, and a DR system offsite 120 miles away:

 

We don't run SIMS, but the bulk of our servers are virtualized, massive project 5 years ago. We run 2 R710 hosts, that were originally standalone 2008R2, now they are clustered 2012R2, both with 140ish GB RAM and dual Xeons and connected to a Equallogic SAN with 21TB of which 14TB is useable.

 

This is backed up by System Center DPM 2012R2 to a physical server with a PERC connected drive shelf, it backs up the complete VM's as well as drives and exchange mailboxes, this is then replicated to another VM running DPM 2012R2 on our DR Host at another site(over 100MB leased line), which is a single R710 with a PERC connected drive-shelf(8TB in a JBOD of 2TB drives) and 70GB of RAM. This does nothing 99% of the time apart from running the single DPM VM.

 

In the case of a DR scenario, we can instruct DPM to restore the VM's in storage to an alternate location, the VM host at the DR site, and then bring them back up, so we can keep going. We decided we could survive for a day, so we dismissed expensive redundant leased lines and instant failover solutions. Taking restore time into account, actually our email and MIS and critical servers will be restored in half a day, but if we had truly gone dark at HQ, finding somewhere else to perch will take longer!

 

We test this, once every 6 months, where we restore the servers to a sandbox(isolated VM network basically) along with a basic DC, and haven't had any problems yet, email, MIS, moodle etc all useable once restored and IP's changed. In a real situation, we would also have to take DNS into account and swap IP's to the new site(we have 30 statics on our leased line at each end).

 

I have a shiny new R730XD with 12TB of storage sitting at the DR site ready to replace the R710 there, as it now past the 7 year ProSupport limit. Just need to find time to visit and build it up. Not very expensive either