Hi guys,
I currently run a few VMs on a 2 node VMware ESXi 5.5 cluster.
Both machines have same specs except for storage ;
Lenovo RD440 (2 sockets but only one populated)
CPU : 1x Intel Xeon E5-2420 v2 @ 2.20Ghz (6 cores)
RAM : 64 GB (all DIMM slots are used)
Production Server storage : 4 x 240GB SSD in RAID5 + 1 single SSD (for a particular hungry ubuntu, MySQL, java calculation node and java app server)
(only 3 HDD bays available)
Replica server storage : 4 x 1TB 7.2K in RAID5
(4 HDD bays available)
VMs are replicated using veeam once a day.
Plus I also use Veeam to backup machines weekly on a NAS.
I'm running out of space on production server. Plus performance is not that good in particular for the ERP server running MySQL, plus java app server and java calculation node.
I run PVE at home since a few years and got really interrested in deploying it in a production environment.
What I need ;
- more storage space
- more performance
- efficient replication between the production server and DR (ZFS seems the way to go)
- manual failover for now
My plan is the following :
1. Buy a refurbished server
2. Install PVE 5.3 and build a RAID10 zfs pool (8x SSD or combo similar to #4 below but with bigger SSDs)
3. Create a pve cluster
4. Add a 2nd node by converting one of the RD440 (maybe with 2x 240SSD + 4x 1TB 7.2K, might use SSD for ZIL + L2ARC)
5. Setup replication between the 2 hosts like each 15 min
6. I will manually start servers on node #2 if node #1 (production server) fail
Down the road, I might also use the 2nd RD440 to act as a management station or even a 3rd node to the cluster for quorum to eventually go HA.
(But I'm not there yet)
What do you think of the following server config ?
(is it too much? or not enough to last a few years since it's still 12th generation 2013 machine)
I buy a lot of refurbished Dell to setup as workstations but never did that for servers)
DELL PowerEdge R820 Server 2.60Ghz 32-Core 192GB 8x 960GB SSD (maybe from techmikeny)
CPU : 4 x 2.60Ghz E5-4650L Xeon Eight Core
RAM : 24 x 8GB PC3L-10600, [192GB Total]
HDD : 8 x 960GB SSD SATA 2.5" 6GB/s
PSU : 2 x DELL 750W Plat
RAID : PERC H710 (I think I will have to switch it for H310 since H710 doesn't support HBA or IT-mode which is needed for ZFS)
Would it be best with current config of 8x SSD in ZFS RAID10 or performance might be just as good with 2x SSD (with ZIL+L2ARC) + 6x 7.2k or 10K SAS spinning drives ?
I mean a more efficient space/cost ratio ?
Other Dell servers : R420, R610, R620, R710 and even R720 seems expensive and not built for the long run.
Maybe I'm mistaken.
Any input would be greatly appreciated.
Thanks.
I currently run a few VMs on a 2 node VMware ESXi 5.5 cluster.
Both machines have same specs except for storage ;
Lenovo RD440 (2 sockets but only one populated)
CPU : 1x Intel Xeon E5-2420 v2 @ 2.20Ghz (6 cores)
RAM : 64 GB (all DIMM slots are used)
Production Server storage : 4 x 240GB SSD in RAID5 + 1 single SSD (for a particular hungry ubuntu, MySQL, java calculation node and java app server)
(only 3 HDD bays available)
Replica server storage : 4 x 1TB 7.2K in RAID5
(4 HDD bays available)
VMs are replicated using veeam once a day.
Plus I also use Veeam to backup machines weekly on a NAS.
I'm running out of space on production server. Plus performance is not that good in particular for the ERP server running MySQL, plus java app server and java calculation node.
I run PVE at home since a few years and got really interrested in deploying it in a production environment.
What I need ;
- more storage space
- more performance
- efficient replication between the production server and DR (ZFS seems the way to go)
- manual failover for now
My plan is the following :
1. Buy a refurbished server
2. Install PVE 5.3 and build a RAID10 zfs pool (8x SSD or combo similar to #4 below but with bigger SSDs)
3. Create a pve cluster
4. Add a 2nd node by converting one of the RD440 (maybe with 2x 240SSD + 4x 1TB 7.2K, might use SSD for ZIL + L2ARC)
5. Setup replication between the 2 hosts like each 15 min
6. I will manually start servers on node #2 if node #1 (production server) fail
Down the road, I might also use the 2nd RD440 to act as a management station or even a 3rd node to the cluster for quorum to eventually go HA.
(But I'm not there yet)
What do you think of the following server config ?
(is it too much? or not enough to last a few years since it's still 12th generation 2013 machine)
I buy a lot of refurbished Dell to setup as workstations but never did that for servers)
DELL PowerEdge R820 Server 2.60Ghz 32-Core 192GB 8x 960GB SSD (maybe from techmikeny)
CPU : 4 x 2.60Ghz E5-4650L Xeon Eight Core
RAM : 24 x 8GB PC3L-10600, [192GB Total]
HDD : 8 x 960GB SSD SATA 2.5" 6GB/s
PSU : 2 x DELL 750W Plat
RAID : PERC H710 (I think I will have to switch it for H310 since H710 doesn't support HBA or IT-mode which is needed for ZFS)
Would it be best with current config of 8x SSD in ZFS RAID10 or performance might be just as good with 2x SSD (with ZIL+L2ARC) + 6x 7.2k or 10K SAS spinning drives ?
I mean a more efficient space/cost ratio ?
Other Dell servers : R420, R610, R620, R710 and even R720 seems expensive and not built for the long run.
Maybe I'm mistaken.
Any input would be greatly appreciated.
Thanks.