Agree with what Bob is saying that you should normally set remote storage via host label.
In terms of overriding the setting via system helm-override-update command. If the configuration is set as a per host override, then you have to override it
in the overrides/nova_compute/hosts section of the nova config.
Gerry
From: Church, Robert [mailto:Robert.Church@windriver.com]
Sent: Monday, September 09, 2019 12:31 PM
To: Friesen, Chris; starlingx-discuss@lists.starlingx.io
Subject: Re: [Starlingx-discuss] Fail to change config of nova-compute container
As Chris mentioned, we are using a host label to drive this configuration on a per worker basis.
If you want to enable this configuration for a specific openstack worker:
$ system host-lock compute-0
$ system host-label-assign compute-0 remote-storage=enabled
$ system host-unlock compute-0
If you look at the nova plugin, this configuration is set via a per-host override based on the host label.
Your nova.yaml is setting the default configuration not the per-host configuration.
I suspect that if we weren’t setting the per-host override for this value explicitly in the nova plugin based on the presence/absence of the label, then this value would
apply to all nova pods across all the compute workers.
Bob
From: Chris Friesen <chris.friesen@windriver.com>
Date: Monday, September 9, 2019 at 11:01 AM
To: "starlingx-discuss@lists.starlingx.io" <starlingx-discuss@lists.starlingx.io>
Subject: Re: [Starlingx-discuss] Fail to change config of nova-compute container
In this case rather than explicitly overriding nova.conf I think the "proper" way to get rbd storage is to set the "remote-storage" label on the node. This allows the rbd_pool and rbd_cev_conf settings to be configured properly as well.
Chris
On 9/9/2019 9:51 AM, Chris Friesen wrote: