My understanding for your questions.

- [Greg] Do we want Ironic to be an OPTIONAL service in StarlingX ?

[Mingyuan] I’m going to add ironic to stx-openstack and enable it by node label, as Brent said.

- [Greg] Where do we want Ironic Conductor running ?

[Mingyuan] I suppose ironic is running on controllers since ipmi connects to oam network.

- [Greg] Is it possible to have an HA deployment of Ironic Conductor in vanilla upstream ironic ?

[Mingyuan] ironic conductor HA is supported in vanilla upstream ironic, ironic conductor pod will run on both controllers. I need further investigation of this.

Ø “There can be multiple instances of the conductor service to support various class of drivers and also to manage fail over. Instances of the conductor service should be on separate nodes.”

Ø https://docs.openstack.org/ironic/latest/install/get_started.html#deployment-architecture

- [Greg] How do we set it up such that system can manage both VM Instances and Bare Metal Instances ? will/can these be setup as part of starlingx system overrides ?

[Mingyuan] Sounds good with host aggregate. For overrides, let me figure this out.

- [Greg] I don’t understand why doing testing of ironic on hardware will take a long time.

[Mingyuan] Bare metal deployment manually of starlingx usually takes more time than that in vm. In addition, to test functionality of ironic, more ironic nodes are needed for installing tenant instances, which obviously takes more time than installing the vm instances.

Mingyuan

From: Waines, Greg [mailto:Greg.Waines@windriver.com]
Sent: Thursday, March 14, 2019 20:56
To: Qi, Mingyuan <mingyuan.qi@intel.com>; starlingx-discuss@lists.starlingx.io
Cc: Rowsell, Brent <Brent.Rowsell@windriver.com>; Miller, Frank <Frank.Miller@windriver.com>; Xie, Cindy <cindy.xie@intel.com>; An, Ran1 <ran1.an@intel.com>
Subject: Re: [Container] Ironic service containerization

In-lined comments,

Greg.

From: "Qi, Mingyuan" <mingyuan.qi@intel.com>
Date: Thursday, March 14, 2019 at 4:45 AM
To: "starlingx-discuss@lists.starlingx.io" <starlingx-discuss@lists.starlingx.io>
Cc: "Rowsell, Brent" <Brent.Rowsell@windriver.com>, "Miller, Frank" <Frank.Miller@windriver.com>, Greg Waines <Greg.Waines@windriver.com>, "Xie, Cindy" <cindy.xie@intel.com>, "An, Ran1" <ran1.an@intel.com>
Subject: [Container] Ironic service containerization

Hi folks,

I’m starting the story of containerize ironic service, I’d like to add below tasks to story https://storyboard.openstack.org/#!/story/2004760 in order to enable ironic service.

Add ironic chart to stx-openstack armada manifest.
[Greg]
Do we want Ironic to be an OPTIONAL service in StarlingX ?
My initial thoughts are yes ... thinking most StarlingX deployments will not use it, so for utilization reasons it would be better not to even run it.

So if we are making Ironic OPTIONAL and disabled by default,
how is this achieved ?
e.g.
      - does it have a separate stx-ironic-openstack armada manifest ?
      OR
      - Is it included in the stx-openstack armada manifest and somehow(?) disabled by default,
        and can be enabled thru helm-overrides ?

Add ironic chart values to armada manifest as well as static configuration.

Generate ironic overrides

Generate dynamic configuration in sysinv for ironic.
[Greg]
As part of this work I think it would be good to understand (and review with starlingx cores):
    - Where do we want Ironic Conductor running ? Controller or Compute ? Do we have a choice in vanilla upstream ironic ?
    - Is it possible to have an HA deployment of Ironic Conductor in vanilla upstream ironic ?
    - How do we set it up such that system can manage both VM Instances and Bare Metal Instances ?
      i.e. I believe this is done thru host aggregates ?   will/can these be setup as part of starlingx system overrides ?

Greg.

Enable ironic service

Adding ironic chart to armada chart group to switch on ironic service for stx-openstack application.

Should this be default in stx-openstack armada manifest?

Cleanup ironic in sysinv and puppet

I’m not sure whether to add this task here or to a specific openstack component cleanup story.

And speaking of the test, the first choice is to deploy starlingx standard config on bare metal because of the connection to ipmi, but it will take long time for testing. Did anyone get the chance to have ipmi/BMC enabled in vm (or it’s not possible)?

Best Regards,

Mingyuan