[Pkg-iscsi-maintainers] Plans for open-iscsi post Jessie

[Ritesh Raj Sarraf]

Hi Christian,


On Thursday 23 April 2015 05:19 PM, Christian Seiler wrote:
> Hi Ritesh,
> (CC-ing you explicitly because you told me you had trouble receiving
> mails via this list.)
>
> Since the Jessie release is now around the corner, I'd like to do some
> work on the package for Stretch. I'm asking you here whether you have
> any objections to my plan:
>

Awesome. :-)

>  - switch to dh(1) instead of manual calls for dh_*, and override
>    where necessary
>
>    This should make the rules file quite a bit easier, and immediately
>    fix #777602 (dh_installinit call deprecated). Also, future
>    improvements to debhelper will automatically be picked up.
>

Yes. We should transition to the newer tools.

>  - move /etc/init.d/umountiscsi.sh to somewhere else, where it can be
>    called from both an init script and a systemd service file (I don't
>    think it makes sense to make this an init script, it should just be
>    called when stopping the iSCSI service)
>

Fine with me. I think just the lintian (old) scripts may do some minor
complaining. But we can override them.

>  - rework umounting script to fix bugs #765431, #775838
>
>  - systemd (1): install upstream iscsid.{socket,service} for iscsid
>

Yes. This is needed.

>  - systemd (2): create proper open-iscsi.service that just does the
>    mounting, but doesn't start iscsid (leave that to upstream's native
>    service files). The init script still does start iscsid, however.
>

Before the whole systemd thing, in Red Hat, there used to be 2 scripts.
One for starting the daemon, the other for establishing the sessions
(login, logout etc).

I guess with systemd too the behavior may be the same. I haven't looked
at it.

>  - fix bugs #764409 (hardening incomplete), #627908 (typo), #777084
>    (udeb for x32)
>
>  - go through the rest of the bugs and either fix them, mark them as
>    already fixed or close them as wontfix (I'd really like to get the
>    open bug count down to zero)
>

:-)

This is always welcome.

> Finally, I'd also like to properly fix the remote-fs issue I originally
> reported. The issue fundamentally came about because there is something
> conceptually wrong: open-iscsi is installed in /usr (so may depend on
> remote filesystems, such as /usr on NFS, hence Requires-Start:
> $remote_fs in the init script), but is itself responsible for bringing
> up remote filesystems.
>
I agree. It is just that technically iscsi brings the block device,
which may or may not have a file system, and may or may not be automounted.

That is one of the things the event based init have claimed to fix. I
guess once you put up the solution, it will show up better.

> I see two proper fixes for this: move iSCSI from /usr to / (as other
> tools have been moved previously), or Debian decides it will now always
> mount separate /usr partitions in initramfs already, so that the move
> is not necessary (but then the dependency for sysvinit can be dropped).
> I'd therefore ask the release team and on debian-devel what the policy
> for Stretch is going to be in this regard, and depending on the answer
> I might then also update the open-iscsi package to move it to /.
>
> Do you have any objections to the changes I've planned?

I have no objections at all.  :-) Thanks for working on it.

And you are right on about the discussion on debian-devel and with other
concerned parties. init and systemd developers for sure.
>
> Regards,
> Christian
>

Thank you for looking into it.  I look forward to seeing you co-maintain
open-iscsi in Debian.

Ritesh

-- 
Ritesh Raj Sarraf
RESEARCHUT - http://www.researchut.com
"Necessity is the mother of invention."


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: OpenPGP digital signature
URL: <http://lists.alioth.debian.org/pipermail/pkg-iscsi-maintainers/attachments/20150424/88e678b4/attachment.sig>