Lenovo Server RD640 - Raid Controller 710

[sfidencio]

Good night,

1. I have two RD 640 servers with identical configurations, as follows:

2. It has 64GB of RAM.

3. Two 300GB SAS disks

4. RAID OffBoard Controller Card, R710, with 1GB Cache. Supports Raid 0/1/10/5/50/60

5. The customer will purchase an additional 6 disks of 960GB SSD or Mechanical Disks of 1TB each, with a speed of 10k.

6. I thought to do the following:
6.1. Perform a RAID 1 on the controller, on the 300GB disks and install the Proxmox / Debian operating system. In short, the virtualizer Host.

7. Make a RAID 5 or 6, with the other 6 disks (Be it SSD or Mechanical), and make the storage pool for the Virtual Machines (VM's). I will use the normal LVM file system and [NOT ZFS], that is, my RAID will be directly on HARDWARE, and not via SOFWARE, any problem in this regard?

8. After performing the initial configurations, I intend to create a Cluster configuration, and replication of VM's from one node to another. Is it possible to configure my replication in real time, without using ZFS or CephFS?

9. Considering that it is not possible to replicate data from one node to another without using ZFS or Ceph, can I use ZFS or Ceph using my RAID configuration over HARDWARE? or I will have to configure my disks without RAID, in other words, no-RAID, with that it will appear to me in PROXMOX, two disks of 300GB and 6 disks of 1TB individually, that is, separate volume, and then I will have to do the RAID using ZFS? Is the RAID configuration over HARDWARE not more secure than software RAID?

10 . In short, I want you to give your opinion about the best scenario to be implemented with ProxMox, according to the HARDWARE configurations proposed above. They can describe sequentially.

Thanks My Friends!

 

[wolfgang]

Hi,

any problem in this regard?

No, this is fine.
The alternative is you like the capability for snapshots, you can use LVM-Thin.
This is not as fast but has snapshots.

 

[bobmc]

Potentially you could use LVM mirroring for replication
https://www.continuitysoftware.com/...on/using-lvm-mirroring-for-disaster-recovery/

Ceph/ZFS would probably be better for replication but they work better with direct access to the drives. You CAN use JBOD disk mode on a controller but it's not recommended for most situations.

 

[sfidencio]

Good afternoon, Thanks to everyone who answered my questions.

1. I finished the tests in my laboratory environment, using virtualbox, simulating a real environment with 2 two Physical Machines, and a third machine, with Debian Linux, which will be the node fake (qdevice). It worked 100%, "a tear almost got out of my eyes. Lol". I used ZFS with raid 5 to store data from VM's and LXC containers, and 2 disks in ZFS raid 1, for the Operating System.

2. In agreement with the client, it was decided to use 2 physical machines from Lenovo RD640, as I described the configurations above, and to my delight, he had recently purchased a Raspberry PI with 1 Gigalan network input, I asked him install debian on it. it was top !!!


3. I will risk using the latest update of the proxmox, which will give me access to the backup feature, 6.2-15, because the 6.2-4 that is the original ISO compiled, does not have. I understand the risks of using unstable packages, but I believe that these packages hardly have bugs, linux is another life, it is not like Microsoft Windows.


4. Thank you, as soon as I become more familiar with the environment, and get more customers, I will purchase annual support, or encourage IT managers to purchase them.


5. I am working hard on the dissemination and use of the environment here in Brazil, in the state of Goiás.


Regards, Sebastião Fidêncio da Silva Pereira