[ovs-dev] [Backport Request] For branch-2.5, backport rhel-systemd integration

[Aaron Conole]

Russell Bryant <russell at ovn.org> writes:

>> On Wed, Sep 7, 2016 at 11:53 AM, Aaron Conole <aconole at redhat.com> wrote:
>>
>>  Greetings all,
>>
>>  I'd like to request a backport to branch 2.5 of the following commits:
>>
>>    c416eaf8c247225f0ebeb22b6ca5c473e96a53d2 ("ovs-ctl: Remove code for upgrading...")
>>    7fc28c50c0128a0c72853f8f243fd5045bcb9917 ("ovs-ctl: Allow selective start for...")
>>    0cfd47f9dcc3914db8d266ed834d6e2c8fc1a11d ("utilities/ovs-ctl.in: Only ...")
>>    48458307b52d13d37d8f88d91385b732af12939f ("utilities/ovs-ctl.in: Allow ...")
>>    4f6218739ec6ed33b71e5127fe619db13605602d ("rhel/ovsdb-server.service: Rename ...")
>>    84ad120834919c3e0945e3e58e0f96c07efa0316 ("rhel: Improved Systemd Integration")
>>
>>  These commits are part of a refactor of the systemd integration which
>>  fixes a number of ordering issues when starting and stopping the Open
>>  vSwitch service on Red Hat systems.  One particularly nasty example
>>  would be configuring an internal interface through Open vSwitch with an
>>  address, mounting an NFS share from that address, and then attempting to
>>  shutdown or restart (which would result in a roughly 1/2 hour wait time
>>  to stop).
>>
>>  Apologies if this is considered in-appropriate for the 2.5 branch, or if
>>  no new backport requests are accepted on the 2.5 branch.
>
> What's the upgrade impact of this for someone using 2.5.0 and then moving to a version including
> these commits?

It should be minimal - I haven't tested pre-refactor to post-refactor
2.5.0 extensively, but I will try it out a bit and if I run into
problems I will self nak this (by 9/16/2017).  It certainly didn't cause
problems on rhel, and the ovs-ctl changes are either flag-enabled
(meaning those flags must be present), or the one commit which removes
compat upgrade code from 2013.

> It sounds like new feature territory, but you do make a case for it being considered a set of fixes ...

I agree - it straddles a line.  I was hesitant to even ask, but in RHEL
we probably need to backport these anyway, so I made an assumption
(maybe poor) that other systemd distros might use the rhel scripts and
run into this class of issues also.  Good net-izen, and all that :)

-Aaron