Hi,
in case you haven't received any answers yet I thought it might still
help if I share our experiences with the upgrade procedure. It really
was tricky in some parts, but we managed to get both Ceph and
OpenStack up and running. We migrated from ceph jewel to luminous and
OpenStack Mitaka to Ocata, not Pike yet.
Basically, your procedure is correct. This is how we did it:
1. upgraded OS of each ceph server, no issues
2. upgraded ceph packages from jewel to luminous on each ceph server,
no issues
3. upgraded ceph packages from jewel to luminous on all cloud nodes,
no problems
4. upgraded controller to Newton, no problems yet
clients could still work properly if they were in external
networks, because neutron had to be stopped
5. double-upgraded control node OS and Newton to Ocata because
ceph-client caused some troubles
6. upgraded OS of compute nodes last, no issues
this was quite easy since we could live-migrate instances to other hosts
The biggest trouble was caused by the database migration. We had to
manipulate the DB ourselves based on the error messages in the logs,
but eventually it worked.
After the most important services were back online I started to update
the configs according to the warnings in the logs for cinder, nova,
neutron etc.
There were some guides that helped me keeping the right order:
[1] https://docs.openstack.org/nova/latest/user/upgrade.html
[2] https://www.rdoproject.org/install/upgrading-rdo/
Unfortunately, I don't have a detailed step-by-step guide about our
procedures and issues we had to resolve. Since it was kind of time
critical I was focused on resolving the problems instead of
documenting everything. ;-)
I hope this helps anyway if you haven't already managed it. We didn't
have a real downtime of our VMs, at least not the ones in production
use since they work in external networks and are not depending on the
neutron services on the control node. Being able to live-migrate
instances was also quite helpful. :-)
Regards,
Eugen
Zitat von Sam Huracan <nowitzki.sa...@gmail.com>:
Hi OpenStackers,
I'm planning upgrading my OpenStack System. Currently version is Mitaka, on
Ubuntu 14.04.5.
I want to upgrade to latest version of Pike.
I've read some documents and know that Mitaka does not have Rolling
upgrade, which means there will have downtime in upgrade process.
Our system has 3 HA Controllers, all VMs and Storage were put in Ceph.
At the moment, I can list some step-by-step to upgrade:
1. Upgrade OS to Ubuntu16.04
2. Upgrage package in order: Mitaka -> Newton -> Ocata -> Pike
3. Upgrade DB in order: Mitaka -> Newton -> Ocata -> Pike
Do I lack any step? Could you guys share me some experiences fulfil
solution, to reduce maximum downtime of system?
Thanks in advance.
--
Eugen Block voice : +49-40-559 51 75
NDE Netzdesign und -entwicklung AG fax : +49-40-559 51 77
Postfach 61 03 15
D-22423 Hamburg e-mail : ebl...@nde.ag
Vorsitzende des Aufsichtsrates: Angelika Mozdzen
Sitz und Registergericht: Hamburg, HRB 90934
Vorstand: Jens-U. Mozdzen
USt-IdNr. DE 814 013 983
_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
