[openstack-dev] [nova] How to debug no valid host failures with placement

Ben Nemec openstack at nemebean.com
Thu Aug 2 15:07:05 UTC 2018



On 08/01/2018 06:05 PM, Matt Riedemann wrote:
> On 8/1/2018 3:55 PM, Ben Nemec wrote:
>> I changed disk_allocation_ratio to 2.0 in the config file and it had 
>> no effect on the existing resource provider.  I assume that is because 
>> I had initially deployed with it unset, so I got 1.0, and when I later 
>> wanted to change it the provider already existed with the default value. 
> 
> Yeah I think so, unless the inventory changes we don't mess with 
> changing the allocation ratio.

That makes sense.  It would be nice if it were more explicitly stated in 
the option help, but I guess Jay's spec below would obsolete that 
behavior so maybe it's better to just pursue that.

> 
>>   So in the past I could do the following:
>>
>> 1) Change disk_allocation_ratio in nova.conf
>> 2) Restart nova-scheduler and/or nova-compute
>>
>> Now it seems like I need to do:
>>
>> 1) Change disk_allocation_ratio in nova.conf
>> 2) Restart nova-scheduler, nova-compute, and nova-placement (or some 
>> subset of those?)
> 
> Restarting the placement service wouldn't have any effect here.

Wouldn't I need to restart it if I wanted new resource providers to use 
the new default?

> 
>> 3) Use osc-placement to fix up the ratios on any existing resource 
>> providers
> 
> Yeah that's what you'd need to do in this case.
> 
> I believe Jay Pipes might have somewhere between 3 and 10 specs for the 
> allocation ratio / nova conf / placement inventory / aggregates problems 
> floating around, so he's probably best to weigh in here. Like: 
> https://review.openstack.org/#/c/552105/
> 



More information about the OpenStack-dev mailing list