Details
-
Question/Request
-
Resolution: Fixed
-
Minor
-
None
-
Lustre 2.12.9, Lustre 2.15.1
-
None
-
x86_64, zfs
-
9223372036854775807
Description
Hi,
short version:
will 2.12 servers and 2.15 clients interop ok?
I don't know if you do interop testing between LTS releases, but it seems like something that could be desirable to make upgrade paths clearer.
longer version:
we have a cluster refresh coming up at the end of the year. currently our clients are rhel7/8 kernels with rhel7 userland and lustre 2.12.x.
lustre servers are rhel7 and 2.12.x, all zfs, all accessed via lnet routers.
the new compute client hardware is planned to be rhel9(alma9/rocky9/whatever) which seems to force us to change to lustre 2.15 as 2.12 doesn't support rhel9.
some lustre servers are being expanded and others decommissioned, but ideally we'd leave those at 2.12 and rhel7 for a while and not have to upgrade everything at once. lnet routers are being added to talk to the new compute.
can we keep the lustre servers at 2.12 and still talk to 2.15 clients ok, or do we have to upgrade the servers to 2.15 at the same time we add 2.15 clients?
if 2.15 on servers is required, then is rhel7 ok, or rhel8 recommended, or ...?
we are zfs 0.8.5 at the moment.
any hints and advice is welcome.
thanks.
FWIW a couple of days ago I briefly booted an alma9 node with git master lustre + 1 rhel9 patch to our 2.12 servers, and didn't notice any problems, but also didn't do much more than boot to root-on-lustre, and ls and df.
cheers,
robin