[lustre-discuss] upgrade process from zfs pool based lustre-2.12 (RH7) to lustre-2.15 (RH8)

[Bernd Melchers]

Hi All,
we plan to update our lustre from 2.12.9 to 2.15.4. We will update the OS from CentOS 7.9 to Alma 8.9, the recommended [1] OS version for 2.15.4. The system setup is based on ZFS, which is at (slightly patched) 0.7.13 and will be updated as
recommended to version 2.1.11.

We did not see a guide or experience report for this update. Any pointers to a document, or any hints about pitfalls to avoid? We were for wondering if it is better practice to leave the zpools at the old level for a system
that has seen its share of crashed in several years of operation, or if "zpool upgrade -a" is the better option prior to start lustre 2.15.

Also we think about using the most recent version of zfs 2.1, which is 2.1.15 and not 2.1.11

The content of our /etc/modprobe.d/zfs.conf is:
options zfs_arc_max=74959079424 zfs_vdev_scheduler=deadline metaslab_debug_unload=1 zfs_vdev_async_read_max_active=16 zfs_vdev_aggregation_limit=1 zfs_vdev_async_write_active_min_dirty_percent=20
zfs_vdev_async_write_min_active=5 zfs_vdev_sync_read_min_active=16 zfs_vdev_sync_read_max_active=16

and /etc/modprobe.d/lustre.conf is:
options lnet networks="o2ib0(ib0),tcp0(eno1)"
options ptlrpc at_min=40 at_max=400 ldlm_enqueue_min=260

Should we change/remove/add some parameters?

Best regards
Bernd Melchers

[1] https://wiki.whamcloud.com/display/PUB/Lustre+Support+Matrix
-- 
Archiv- und Backup-Service | fab-service at zedat.fu-berlin.de
Freie Universität Berlin   | Tel. +49-30-838-55905