Re:[watcher] Compute CDM builder issues (mostly perf related)

li.canwei2 at zte.com.cn li.canwei2 at zte.com.cn
Thu May 16 02:07:29 UTC 2019


On 5/15/2019 4:19 AM, li.canwei2 at zte.com.cn wrote:
> I tried changing nova_client.api_version to a FloatOpt but that gets
> messy because of how things like 2.60 are handled (str(2.60) gets turned
> into '2.6' which is not what we'd want). I was hoping we could use
> FloatOpt with a min version to enforce the minimum required version, but
> I guess we could do this other ways in the client helper code itself by
> comparing to some minimum required version in the code.
> [licanwei]: Maybe we can refer to 
> https://github.com/openstack/watcher/blob/master/watcher/common/nova_helper.py#L714 
> 

I just did this which seems more explicit:

https://review.opendev.org/#/c/659194/

That change leaves the default of 2.56 since the 2.56 code does version 
discovery so it's backward compatible, but I think we can assert that 
you need at least 2.53 because of how the scoped nova CDM code works 
(and to support nova deployments with multiple cells properly).

Also note that 2.53 is pike-era nova and 2.56 is queens-era nova and 
those seem old enough that it's safe to require 2.53 as a minimum for 
watcher in train.
[licanwei]: Because we need to specify the destination host when migration,

at least 2.56 is required.

 https://github.com/openstack/watcher/blob/master/watcher/common/nova_helper.py#L145

-- 

Thanks,

Matt
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-discuss/attachments/20190516/1af320b5/attachment.html>


More information about the openstack-discuss mailing list