Hi all.
Some backround on the environment first:
Four single CPU equipped DELL PowerEdge R620's.
Currently running two single ESXi nodes each with local storage and no HA or replication, one physical file server / Domain Controller and a physical backup server with tape drive.
Almost pure Windows environment with two DC's, an Exchange and a Forefront TMG. Some of that will change, but that's it for now.
I am looking to optimize the usage of these servers and implement as much HA as possible and strongly considering going for proxmox for the virtualization.
I'd like to have the backup server physical to replicate backups to tape for offline/offsite backup to have some DR in case of fires and such.
That leaves me with three servers. I read some stuff in the forum and wiki about HA without shared storage, but some info seems old and outdated.
I am thinking something like the following
- Three clustered proxmox nodes with (close to) instant replication (Ceph?)
- Local storage on Raid10's with the servers PERC Controllers
- Each Server with 4 Gigabit NICs bonded with e.g. LACP, maybe with two switches between them for redundancy
The whole thing shall be online-backuped to the backup server (which is currently running Windows but does not have to), which puts it to tapes fo DR.
Goals are:
- Live migrations for maintenance on nodes
- Failover without data loss in case of hardware defects on a node (does not have to be HA or fully automatic, but the less manual intervention needed the better of course)
I that scenario possible/realistic?
If the above it not possible, I may have the ability to get some older Servers for backup (one as spare) and use all four R620's in a setup for e.g. a two node VE cluster and a two node storage. But for various reasons I'd like to avoid that.
Other recommendations are also very welcome.
If not possbile otherwise, downtimes are to some extend tolerable on hardware failure, loss of data (e.g. between replication times) is not.
Thank you in advance for advice and/or recommendations,
Hannes
Some backround on the environment first:
Four single CPU equipped DELL PowerEdge R620's.
Currently running two single ESXi nodes each with local storage and no HA or replication, one physical file server / Domain Controller and a physical backup server with tape drive.
Almost pure Windows environment with two DC's, an Exchange and a Forefront TMG. Some of that will change, but that's it for now.
I am looking to optimize the usage of these servers and implement as much HA as possible and strongly considering going for proxmox for the virtualization.
I'd like to have the backup server physical to replicate backups to tape for offline/offsite backup to have some DR in case of fires and such.
That leaves me with three servers. I read some stuff in the forum and wiki about HA without shared storage, but some info seems old and outdated.
I am thinking something like the following
- Three clustered proxmox nodes with (close to) instant replication (Ceph?)
- Local storage on Raid10's with the servers PERC Controllers
- Each Server with 4 Gigabit NICs bonded with e.g. LACP, maybe with two switches between them for redundancy
The whole thing shall be online-backuped to the backup server (which is currently running Windows but does not have to), which puts it to tapes fo DR.
Goals are:
- Live migrations for maintenance on nodes
- Failover without data loss in case of hardware defects on a node (does not have to be HA or fully automatic, but the less manual intervention needed the better of course)
I that scenario possible/realistic?
If the above it not possible, I may have the ability to get some older Servers for backup (one as spare) and use all four R620's in a setup for e.g. a two node VE cluster and a two node storage. But for various reasons I'd like to avoid that.
Other recommendations are also very welcome.
If not possbile otherwise, downtimes are to some extend tolerable on hardware failure, loss of data (e.g. between replication times) is not.
Thank you in advance for advice and/or recommendations,
Hannes
Last edited:
