Several questions about roadmap

[Андрей Башлаков]

Hi all!

1) About use case for RBAC: we have hardware engineers in ours DCs, which responsible for mounting new servers in racks, commisionning, testing and replacing faulted hardware. In our scenario they need to prepare hardware and make possible to use it by system administrators. So they need MaaS functions related to testing and commissioning but they not allowed to manage this hardware (acquire, deploy etc).
2) About hooks and polling. Polling is very slow for us, getting machines info via API took up to 30 seconds on 3 hundred machines and we planning to expand total amount of machines up to several thousand in near future. We using our custom software to wrap MaaS API and run test on machines and we need to decrease status update lag (from running test for machine to get updated status for this machine) to several seconds.

Unfortunately, I’m just junior python programmer (and mostly system administrator rather than developer) and not sure that I can contribute to MaaS source. But I will be very glad to se this changes in MaaS.


31 окт. 2017 г., в 11:14, Patrizio Bassi <patrizio.bassi at gmail.com<mailto:patrizio.bassi at gmail.com>> написал(а):



2017-10-31 9:10 GMT+01:00 Mark Shuttleworth <mark at ubuntu.com<mailto:mark at ubuntu.com>>:
On 10/30/2017 10:51 PM, Michael Iatrou wrote:
On Mon, Oct 30, 2017 at 9:16 AM, Андрей Башлаков <a.bashlakov at corp.vk.com<mailto:a.bashlakov at corp.vk.com>> wrote:
Hi, all!

Hello Andrei!

We have several questions about MaaS roadmap:
1) Is there any plans to make role-based user management? Current scheme with admins and users is very constrained. We would like to see ability to add roles that can, for example, only commission and run test on new machines but can’t acquire machines or do anything else;

Yes, we do expect to have RBAC capabilities in future.


2) Is there any plans to implement custom hooks for events, such as change machine state or discovery new machines? We developing internal product that actively using MaaS API and currently we need to run several crawlers for detecting such changes.

How often does the state of the machines change?
How large is the pool of the managed machines?
Any reason why the polling approach is not sufficient?

Polling won't scale very well :)

I think the idea of hooks is sensible. It's not on our roadmap, patches or engineering engagements around that would be welcome.

Mark


+1 to the idea of hooks with filters (hostname/ip address/hw features). When something happens (es. machine spawns) we may want to do something like register in inventory systems, subscribe to backup and monitoring, send an email, notice a bot.....

Thank you

--
Patrizio Bassi


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ubuntu.com/archives/maas-devel/attachments/20171031/7a51baea/attachment.html>