[Openstack-operators] Upgrading from Havana to Icehouse

Jonathan Proulx jon at jonproulx.com
Wed Jul 9 13:29:58 UTC 2014


I'm still in testing phase too, but here are my thoughts.

My plan is to upgrade the controller on both dimensions
Precise->Trusty and Havana->Icehouse.  In my case this will actually
be a cut over between systems not an in place upgrade. So on upgrade
day I'll have a configured and tested Trusty/Icehouse controller
standing by with services down, will shut down  the current
Precise/Havana controller, connect new controller to the live DB store
(an iscsi volume), run migrations & restart services.

For compute nodes I'm only doing the Havana->Icehouse upgrade
initially.  Then over the next 6 months (hopefully less) will do
rolling upgrades from Precise->Trusty using live migration juggling,
so we're read to go for next round of OpenStack upgrades which we'd
plan on doing in early January (for internal reasons mid summer and
right after new year work best for major maintenance here).

I'm about 2 weeks from first round of this and <knock-wood> remain
optimistic </ knock-wood> about the plan, off to run more test
upgrades now.

-Jon


On Tue, Jul 8, 2014 at 7:54 PM, Mathieu Gagné <mgagne at iweb.com> wrote:
> On 2014-07-08 7:36 PM, Xav Paice wrote:
>>
>>
>> Like many people, I'm planning our upgrade from Havana to Icehouse, and
>> just want to check off a few items before taking the plunge.  So far,
>> testing is going well and it doesn't look like a painful upgrade.
>>
>> I was wondering if, among those of us that use Ubuntu, people have been
>> upgrading from Precise to Trusty along with the OpenStack upgrade or
>> treating that as a separate exercise? Are people using Icehouse on
>> Precise and planning to stay that way for long?
>>
>
> My plan is to upgrade to Icehouse on Precise.
>
> Upgrading to Trusty is a separate exercise, especially when you have to
> consider the work required to upgrade compute nodes to Trusty. It will
> surely require a reboot and it can't be done without careful planning with
> the customers.
>
> You could still juggle instances around with live migration but forget about
> upgrading to Icehouse in less than an hour. In our case, it would take
> several days.
>
> I usually try to split maintenance in pieces as small as I can to avoid
> piling up problems and reducing variables.
>
> --
> Mathieu
>
>
> _______________________________________________
> OpenStack-operators mailing list
> OpenStack-operators at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators



More information about the OpenStack-operators mailing list