[rdo-list] [TripleO] Newton large baremetal deployment issues

[Charles Short]

On 03/11/2016 13:30, Justin Kilpatrick wrote:
> It sounds like you're running more into node assignment issues rather 
> than pxe issues though

Hi,

Thanks for the reply. This is an interesting thought. I already assign 
the controllers to specific nodes, but let the compute nodes get 
assigned at random.
I will try pinning the compute nodes and see if that helps.
A 50 node deployment is about to complete. It looks like it will take in 
the region of 6-7 hours to complete!
I will then assign the remaining compute nodes explicitly and run a 
stack update with the extra nodes.
I had previously tried the approach of lots of updates of smaller 
numbers of nodes, but I found that for example adding 10 nodes to a 
deployment of say 30 nodes would also take an age (hours and hours).

I will report back after more testing

Charles



-- 
Charles Short
Cloud Engineer
Virtualization and Cloud Team
European Bioinformatics Institute (EMBL-EBI)
Tel: +44 (0)1223 494205