Agree with this approach. The large-scale test scenario configuration manual is very meaningful.

When the OpenStack deployment scale reaches a certain level (for example, nodes >= 200,500… etc.),

various exception scenarios will occur, rabitmq blocking conditions, and create a batch of servers.

It’s success rate (There will be some configurations to be care. e.g. scheduling policies, rpc wait time, amount of works etc.).

If there is a reference manual, this is very friendly.

On Sat, 3 Aug. 2019, 6:25 am Matt Riedemann, <mriedemos@gmail.commailto:mriedemos@gmail.com> wrote:

I wanted to send this to get other people's feedback if they have particular nova configurations once they hit a certain scale (hundreds or thousands of nodes). Every once in awhile in IRC I'll be chatting with someone about configuration changes they've made running at large scale to avoid, for example, hammering the control plane. I don't know how many times I've thought, "it would be nice if we had a doc highlighting some of these things so a new operator could come along and see, oh I've never tried changing that value before".

I haven't started that doc, but I've started a bug report for people to dump some of their settings. The most common ones could go into a simple admin doc to start.

I know there is more I've thought about in the past that I don't have in here but this is just a starting point so I don't make the mistake of not taking action on this again.

https://bugs.launchpad.net/nova/+bug/1838819

--

> > Thanks,

> Matt

Hi Matt,

My name is Joe - docs person from years back - this looks like a good initiative and I would be up for documenting these settings at scale.

Next step I can see is gathering more Info about this pain point (already started :)) and then I can draft something together for feedback.