Upgrade: Juno -> Pike (or maybe later)
Hey -- I've run a number of OpenStack clouds from multiple vendors for several years, now. But I've *never* done an upgrade. I've read some of the docs, but they don't really make it clear (at least, the ones I've thus-far found): * If going from Juno to something pretty recent is even feasible * If I have to do intermediate steps along the way * How likely it is to "just work:" complexity, common issues, etc.
The fact that there's a clearly codified fallback procedure, while an inherently good thing to have, does concern me some.
Any insights would be much appreciated.
Thanks!
-Ken
On 12/12/2019 3:50 PM, Ken D'Ambrosio wrote:
Hey -- I've run a number of OpenStack clouds from multiple vendors for several years, now. But I've *never* done an upgrade. I've read some of the docs, but they don't really make it clear (at least, the ones I've thus-far found):
- If going from Juno to something pretty recent is even feasible
- If I have to do intermediate steps along the way
- How likely it is to "just work:" complexity, common issues, etc.
The fact that there's a clearly codified fallback procedure, while an inherently good thing to have, does concern me some.
Any insights would be much appreciated.
Oh man.
Well, probably start here [1].
And watch some of these [2].
This thread should probably be tagged with [ops] because it's a long-standing thing that several people have done but I'm not sure how well any of it is documented and that wiki might be old.
BTW, if your clouds are from vendors why aren't the vendors doing your upgrade?
[1] https://wiki.openstack.org/wiki/Fast_forward_upgrades [2] https://www.openstack.org/videos/search?search=fast%20forward
On Thu, Dec 12, 2019 at 4:53 PM Ken D'Ambrosio ken@jots.org wrote:
Hey -- I've run a number of OpenStack clouds from multiple vendors for several years, now. But I've *never* done an upgrade. I've read some of the docs, but they don't really make it clear (at least, the ones I've thus-far found):
- If going from Juno to something pretty recent is even feasible
Stuff will probably break, you're talking about upgrading from 11 releases.
There's just SO much that happened in that period of time & transitions.
- If I have to do intermediate steps along the way
Many, placement extraction, migration to cells v2 are two big ones that come to mind
- How likely it is to "just work:" complexity, common issues, etc.
Quite honestly, the complexity is very high of what you're trying to accomplish and I would seriously consider redeploying your clouds on new hardware and migrating workloads at this point.
The fact that there's a clearly codified fallback procedure, while an inherently good thing to have, does concern me some.
Any insights would be much appreciated.
Thanks!
-Ken
Hi, I agree here, maybe a forklift upgrade might be easier. As Mohammed pointed out the placement and cells are one of the big ones. If anything, you would need to upgrade just pre placement and pre cell and then use the upgrade path there to overcome that hurdle and then move forward. With 11 (at least) releases , it’s gonna be rough. I would argue, carefully with the amount of changes, you are trying to migrate windows 3.11 to windows 10... It probably can be done but (aside the 32 bit etc:)) ... wow
If you need to upgrade on existing hardware, why not consolidate some instances and free up some nodes and build the environment then up?
Good luck!:)
On 13. Dec 2019, at 0.41, Mohammed Naser mnaser@vexxhost.com wrote:
On Thu, Dec 12, 2019 at 4:53 PM Ken D'Ambrosio ken@jots.org wrote:
Hey -- I've run a number of OpenStack clouds from multiple vendors for several years, now. But I've *never* done an upgrade. I've read some of the docs, but they don't really make it clear (at least, the ones I've thus-far found):
- If going from Juno to something pretty recent is even feasible
Stuff will probably break, you're talking about upgrading from 11 releases.
There's just SO much that happened in that period of time & transitions.
- If I have to do intermediate steps along the way
Many, placement extraction, migration to cells v2 are two big ones that come to mind
- How likely it is to "just work:" complexity, common issues, etc.
Quite honestly, the complexity is very high of what you're trying to accomplish and I would seriously consider redeploying your clouds on new hardware and migrating workloads at this point.
The fact that there's a clearly codified fallback procedure, while an inherently good thing to have, does concern me some.
Any insights would be much appreciated.
Thanks!
-Ken
-- Mohammed Naser — vexxhost
D. 514-316-8872 D. 800-910-1726 ext. 200 E. mnaser@vexxhost.com W. https://vexxhost.com
Ken D'Ambrosio wrote:
Hey -- I've run a number of OpenStack clouds from multiple vendors for several years, now. But I've *never* done an upgrade. I've read some of the docs, but they don't really make it clear (at least, the ones I've thus-far found):
- If going from Juno to something pretty recent is even feasible
- If I have to do intermediate steps along the way
- How likely it is to "just work:" complexity, common issues, etc.
The fact that there's a clearly codified fallback procedure, while an inherently good thing to have, does concern me some.
Any insights would be much appreciated.
OVH did migrate all its regions from Juno to Newton, you should be able to find presentations on that from past summits videos.
Similarly, Verizon media / Oath did migrate from Juno to Ocata. James Penick did presentations and some blog articles about it.
But yes, as others said, that would certainly not be the first upgrade I ever attempted. Juno pre-dates most of the work on simplifying upgrades, so it's a painful starting point.
participants (5)
-
Florian Rommel
-
Ken D'Ambrosio
-
Matt Riedemann
-
Mohammed Naser
-
Thierry Carrez