Hi, everyone.
I’m looking into implementing a iSCSI HA redundant storage and after doing some research I noticed that many of the provided documentations are almost 10 years old.
One of the solutions I came across is DRBD + HAProxy + any iSCSI target software.
Would it work and would it be an Up-to-date solution. Or are there any better solutions out there.
I would like to do it on a Debian or a OpenSUSE system. Witch one would you prefer for a setup like this?
Looks like you can run iSCSI on top of GlusterFS. I’ve never done it, but that’s where I would start if I were trying to accomplish HA/distributed iSCSI.
https://gluster.readthedocs.io/en/latest/Administrator%20Guide/GlusterFS%20iSCSI/
I understand that in this setup I would have 2 Gluster servers that are redundant and a iSCSI target that mounts the Gluster disk and creates a iSCSI LUN on it.
The problem is that the communication between the Gluster servers and the iSCSI target would be in file level not in block level. So in this setup I could just use it as a CIFS share.
Thought about Ceph?
1 Like
Actually no. I will take a look at it right away,
On a smaller scale you could just raid your iscsi volumes (same writehole caveats as physical disk raid apply).
This is an idea… better have a dedicated network for the traffic, IMO
As far as the initiator goes, I know FreeBSD has multipath support for SCSI disks (and I would be floored if Linux doesn’t). I’m not sure what sort of support you are looking for on the target…
That is not quite what the OP is wanting, but redundant paths to the storage is also always recommended. He wants redundant storage at the disk hardware level.
Ok, I wasn’t sure if that was the case because keeping disks on separate servers in sync is serious business.
It’s been a while since I’ve looked at Ceph but that seems like a reasonable solution. Another possibility is DragonflyBSD’s HAMMER filesystem, which supports streaming replication across multiple nodes.
https://www.dragonflybsd.org/hammer/