ZFS: how to enable I-T mode on LSI 9260-8i?

Added by maruti jonnala almost 4 years ago

the config menu (web-boot) option only has RAID-0, 5, 6 for virtual disk groups.....how to enable passthrough or initiator-target mode for ZFS RAIDZ?

does it need a firmware upgrade...

current version is 2.0x

Replies (2)

RE: ZFS: how to enable I-T mode on LSI 9260-8i? - Added by Robin Axelsson almost 4 years ago

I have now learned that JBOD mode can mean two different things. I don't like that they mix the concepts like this but here it goes (according to http://en.wikipedia.org/wiki/JBOD#JBOD):

JBOD, meaning "Just a Box Of Disks" or "Just a Bunch Of Drives" is used to refer to two distinct concepts:

  • all disks being independently addressed, with no collective properties – each physical disk, with all the logical partitions each may contain, being mapped to a different logical volume: just a bunch of disks.
  • concatenation, where all the physical disks are concatenated and presented as a single disk.

And we certainly do not want the latter for a ZFS pool. I'm also very sceptical that a HBA allegedly supporting JBOD implies that it also supports the former concept (unless it is explicitly specified).

As for the 9260-8i, there currently is no IT-firmware for it, and I do not know of any plans to release such firmware anytime soon. If you would like to pass disks directly through to the OS I would recommend the LSI SAS 9211-8i. The point of passing the disks directly to the OS is to make sure that all errors that may occur get passed to the OS and not get obfuscated by the hardware's internal error handling.

But if you already have a 9260 and are unable to swap it for a 9211 the best option you have at disposal is to configure each drive on it as RAID0 with only one disk per stripe. It's not optimal as it tends to cripple the capabilities of the OS to detect errors in the hardware, but at least it works.

RE: ZFS: how to enable I-T mode on LSI 9260-8i? - Added by Richard Elling almost 4 years ago

Also, the 9260 (or similar) are not suitable for shared-disk clusters because they
hide the disk from the OS even when using RAID-0. This means that you can physically
connect a shared disk to two 9260s, but the host OS will not recognize the disk as
shared. But more importantly, the 9260 can cache data that cannot be failed over to
the other node in the cluster. Save yourself some money, go with a simpler HBA.