Been reading about this subject but its info overload, was hoping someone could point me in the right direction as we're on a tight timescale.
To start with we will have 4 HP 685c blades with an EVA 8000 SAN.
There will be roughly 20 VMs per / blade
I guess there are 3 main questions
1)Disater recovery - what do people use. My thinking was to snap clone the VMFS LUN(S) - we could then present these to another ESX server (using re signaturing). But then I thought how do we get this off onto our SAN attached IBM TS 3310 (6 drive tape library) , or is there an easier route to do this, eg, is it easier just to backup all the individual vm's.
I also read about re registering the VMS when representing the VMFS LUN to a new ESX server - has anyone done this ?
2)VM level backups - any thoughts on these ?
3)file level backups - any thoughts ?
Please be gentle with me as this is the first piece of VMware work I've done,
We are currently in the planning stages of DR. We plan to operate 2 VM farms on 2 different subnets in 2 different locations. The plan is to leave enough head room in each farm to be able to run all VMs from both farms on one in a DR scenario. Furtheremore, we will split the running VMs between facilities. This will alllow for quicker failover as only half the VMs will need to be moved at once.
We plan to classify Tier 1 apps (Exchange, SQL, File) and use a product like XOSoft WANSync to replicate the app to another VM in a remote location. This will give us HA between facilities. Failover would take minutes.
For the rest of our VMs (which will not be as heavily used as the Tier 1 apps) we plan on replicating either at the SAN or VMFS level. In a DR scenario we will have our Network guys enable the subnet at the remote facility and then we will start the VMs at the DR site.
Internally we will us HA/DRS to prevent against physical host failure.
This is the current plan. We have not begun testing yet, so I do not know of any issues this plan may encounter.
For our backups at the VM level we us esXpress ( http://www.esxpress.com). There backups run as Virtual Backup Appliances (VBAs) in the virtual space. They can make delta backups along with fulls plus many other features.
We send our backup to a central backup server and then cut tapes from there. For DR we are still just using our tapes at the DR site.
On file level backups think it depends on your specific VMs and if there are business needs there where you need that level of backups. Think at minimum do the vm level (vmdk) backups and then implement a file level backup strategy to compliment the VM level where needed.
Good luck with your new Vmware environment.
there are soooooooooo many options.
there is VCB (shudder at the thought of it..!) which will allow you to go from your running VMs direct to your Tape library via a win2k3 SAN attached "proxy" server. VCB integrates with EMC Networker, TSM, CommVault, Netbackup/ BackupExec and ESXRanger (from Vizioncore).
VizionCore itself will do an admirable job in providing VM backups but only to disk, you'd then use existing Backup solution to push to tape.
alternatively there are some really good console level scripts that will also backup straight to disk (search the forums for VMBK and VisBu). I have customers using both of these options to good effect.
if you have the infrastructure then storage level replication will provide an affective DR solution, I don't know how the EVA replication functions but in an EMC SAN we'd mirror the disk with the VMDKs and then snap the remote mirror, assign it to the Backup server and push it to tape.
or you go with agents in your VMs but beware, it is so easy to deploy a new VM that i have seen customers go this route and end up blowing a large chunk of the initial savings on licensing.
it all comes down to:
a) your RTO (recovery time objective)
b) what you actually need
c) what you can afford.
"...but in an EMC SAN we'd mirror the disk with the VMDKs and then snap the remote mirror, assign it to the Backup server and push it to tape."
Would not you have to make sure that all the VMs on the mirrored LUNs that you are going to snap© are in a consistent state?
Otherwise you may be just copying invalid information, correct?
Lots of options as the othere have mentioned. RDM ( instead of vmfs/vmdks ) + SAN tools is a good option if you have a SAN in both locations.
Tools like express and esxReplicator are great tools for replicatiing virtual machines between locations of distance where the SAN is not an option ( no SAN, too expensive, etc ).
VMware VCB + your favourite backup client are good options for doing local backups. These could be good enough for DR if you send the backups off site and your target recovery time is in the days - weeks time frame and not ours
file level backups can put a good bit of load on your virtual machines although if your doing this at night and your vms are otherwise idle it may not be a big deal. As a DR solution on it's own it isn't that complete and would require more effort at recovery time and push out those recovery times.
Ok this is probably more of an advertisement for VMWORLD 2007, but:
We will be presenting how Florida Hospital is doing DR using vmware. We are currently replicating our luns to a physically remote location. We have already gone through 1 test and it worked like a champ!