iSCSI storage and replication

T

tjarcoboerkoel

New Member
Sep 30, 2023
10
0
1
Hi All,

I'm struggling with proxmox storage, replication and HA.
Does this make sense? I need replication of VMs from SAN1 to SAN2 and I thought it could be achieved by SEPH, but it seems that SEPH cannot use iSCSI storage.

I was thinking about sloution1:

HA_proxmox.jpg


Or solution2:
HA_proxmox2.jpg

But With the limitations of replication and CEPH I'm bumping into walls now. Could someone give me some advice how to approach?

Best regards,
 
Ceph is meant to work with local disks, not SAN. You can force it into Ceph but you would be defeating its purpose.
Ceph is also not a replication technology - its a distributed storage technology. The two have different goals and limitations.

The only built into PVE true replication technology is based on ZFS. ZFS is not a cluster aware filesystem and is generally not meant to be used with SAN, but again you can if you really want to.

You didnt say what is the underlying need for replication between two SANs. Sometimes the backend storage has LUN to LUN replication.
In summary, there is nothing "pretty" out-of-the-box wizardly, today, that can do what you described at high level.

Blockbridge : Ultra low latency all-NVME shared storage for Proxmox - https://www.blockbridge.com/proxmox
 
Hi bbgeek17,

Thnx for your reply,

Well both SANs will probably be on two different physical locations on site. They have no redundant hardware, so no redundant CPU and no redundant power supply. We don;t want it either because a burned PSU would also mean a destroyed SAN (we've seen that before).
And if a server fails, VMs failover to other servers. Thats something we have running now (with 1 SAN).

We try to get something near Windows failover cluster like failover, so data stored on the first SAN and replicated to the 2nd. If the first fails, then failover to the second (without automatic fall back).

There is an iSCSI replication on SAN level available, but on server (service) level would be nice(r).
 
LnxBil said:
Can you elaborate?
Click to expand...
A burned out PSU in a redundant PSU-SAN will kill your SAN completely. I'm not talking about a normal PSU failure which will cause a regular power failure, but a burned PSU will destroy your SAN permanent (acids, carbon particles, overheating and in rare cases flames).

And if one device in the cabinet burns out on its PSU, all components in the cabinet are contaminated. So we have chosen for two cabinets, but we need some synchronization / mirroring of the data.

You can synchronize two CEPHs with each other using RBD Mirroring. But as noted, you can't do this via the PVE GUI.
Click to expand...
I'll look into this, this might be something. The extra sync data on the network is no problem, we have 10gb aggregation switches for the data-paths.
 
tjarcoboerkoel said:
A burned out PSU in a redundant PSU-SAN will kill your SAN completely. I'm not talking about a normal PSU failure which will cause a regular power failure, but a burned PSU will destroy your SAN permanent (acids, carbon particles, overheating and in rare cases flames).

And if one device in the cabinet burns out on its PSU, all components in the cabinet are contaminated. So we have chosen for two cabinets, but we need some synchronization / mirroring of the data.
Click to expand...
Thank you and very informative. Can this not happen to any PSU? I don't see any "special case" for SANs, those are just ordinary PSUs.

I've seen Lampertz safe rack technology that separates racks from each other just for the purpose of separating fire compartments.
 
You must log in or register to reply here.
Top Bottom