Hi Steve,
Thanks for your kind words and very interesting email.
I will take a closer look (read test) your patches as soon as I can, and
if they solve what you mentionned without ill effects, will push them
upstream. It is true that lio-utils is now deprecated, in the sense that
new developments should not use it, but you are right in saying that it
is still out there. As a consequence it should receive bugfixes when
possible. We are even discussing with the debian folks if we should keep
it around even after we start shipping 3.x, to provide support for
legacy installs.
Regarding your use-case, for having deployed myself countless
DRBD+pacemaker iSCSI clusters in the wild, I understand the pain.
As you guessed, I would prefer people to use something else than
lio-utils for that, especially now that the 3.x series with the config
API is out.
With 2.x, what I tend to do in my own installs is to write ad-hoc RA
scripts in python, with minimal parametrization, using the rtslib API,
but also doing stuff like reloading PR metadata and such. Again, these
are very ad-hoc, I never got around to writing a generic RA that I could
publish.
However, with 3.x and the config API, I guess this is something I should
look into. What I have in mind is something very simple and robust,
leveraging the config API: instead of passing all the target
configuration to the RA, simply point it to a configuration snippet that
would be loaded on start and unloaded on stop, without touching the rest
of the running configuration (as long as there are no name/WWN clashes
between target objects for two separate resources).
As someone who got his hands dirty doing some pacemaker clusters, what
do you think of this approach? Would you adopt it? Of course, if syncing
the config snippets across the cluster nodes is a pain, I guess one
could just inline their contents as RA parameter. That would bloat the
CIB a little more, but it should not be a problem, corosync has now
taken over heartbeat and everyone should be able to sync large CIBs if
needed.
As a side note, battling with OCF and wrong RA return codes is a pain I
experienced oh-so-often that I feel bad that lio-utils did that to you
;-) Sorry!
Best Regards,
--
Jerome Martin
--
To unsubscribe from this list: send the line "unsubscribe target-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
