Date
1 - 4 of 4
BOSH multi-zone resurrector
|
Aleksey Zalesov
Hello!
I have a question about resurrection algorithm BOSH uses in multi zone deployment. Suppose we have three availability zones: z1, z2 and z3. We have data/0 vm in z1 with resurrection enabled. Suppose now that z1 goes down. Will resurrector start new data/0 instance in another zone? If yes, how will it chose between z2 and z3? What if BOSH director itself is deployed to the z1? Aleksey Zalesov, Altoros -- View this message in context: http://cf-bosh.70367.x6.nabble.com/BOSH-multi-zone-resurrector-tp526.html Sent from the CF BOSH mailing list archive at Nabble.com. |
|
|
|
Dmitriy Kalinin
Currently Director will bring back VMs in the same AZ as configured in the
manifest because resource pool is configured to point to that specific AZ. If Director is in the affected zone you will have to redeploy Director to a different AZ and restore DB from a backup or point to external DB that it was using. On Mon, Jul 27, 2015 at 12:28 PM, lexsys <aleksey.zalesov(a)altoros.com> wrote: Hello! |
|
|
|
John Wong
Hi also interested in this topic.
Is there a plan to be able to fallback to other zones? This seems like a good capacity management but from automated recovery this seems like chore. What do you suggest? Thanks. John On Fri, Jul 31, 2015 at 6:28 PM, Dmitriy Kalinin <dkalinin(a)pivotal.io> wrote: Currently Director will bring back VMs in the same AZ as configured in the |
|
|
|
Dmitriy Kalinin
We are currently working on adding first class concept of AZ to bosh:
toggle quoted message
Show quoted text
https://github.com/cloudfoundry/bosh-notes/blob/master/availability-zones.md Once that's complete, we can definitely think more about how that should affect the resurrector (I've added tbd item for this). Tricky part is about instance with persistent disks since they cannot be moved to a different AZ. For instances that do not have persistent disks I could see us making a choice of moving them to a different AZ. Few questions arise out of that: - when should we try a different AZ? after a few failures in the assigned AZ? - should next bosh deploy put the instance back into a correct AZ.(and if not that could mean that over time your whole deployment slowly moves into a single AZ) - what to do with instances with persistent disks? should we just throw away the disks and move? What do you think? On Tue, Aug 4, 2015 at 2:50 PM, John Wong <gokoproject(a)gmail.com> wrote:
Hi also interested in this topic. |
|
|