[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

"openstack server list" takes 30s

On 9/10/20 21:15, Tony Liu wrote:
> Hi,
> I built a Ussuri cluster with 3 controllers and 5 compute nodes.
> OpenStack CLI ran pretty fast at the beginning, but gets slower
> over time along with increased workloads. Right now, it takes
> about 30s to list 10 VMs. The CPU, memory and disk usage are on
> those 3 controllers are all in the range. I understand there are
> many API calls happening behind CLI. I'd like to figure out how
> this 30s is consumed, which call is the killer.
> Any guidance or hint would be helpful.

To see the individual calls made by OSC and troubleshoot the reason for 
the slowness, you can use the --debug option:

$ openstack --debug server list

Besides that, there is something that I know of that can be slow: the 
flavor and image name lookup. This happens if you have a Large Number of 
flavors and/or images. There are a couple of options you can use to make 
this faster:

$ openstack server list --name-lookup-one-by-one

This will make OSC lookup flavors and images only once per unique 
flavor/image in the list and uses the cached value for subsequent 
appearances in the list [1]. I think this should have been made the 
default behavior when it was introduced but it was nacked and accepted 
only as opt-in.

The other option skips name lookup altogether and shows UUIDs only 
instead of names [2]:

$ openstack server list --no-name-lookup

Hope this helps,