I'm not exactly sure at the moment what I'll need to get this race
condition to expose itself. Let me give it some thought and take a
closer look at the logs.
On 06/01/2016 09:41 AM, Andreas Hasenack wrote:
> We don't use deployer or amulet. Landscape talks to the juju api
> natively.
>
> We "reserve" machines beforehand via add-machine (both machines and
> containers) after bootstrap, and then we deploy services onto them. We
> also use the juju run trick twice to wait for relations to settle.
>
> There is no simple way to emulate landscape's deployment, you will have
> to use it yourself I guess, or figure out the ordering of events via the
> log files.
>
> If these logs are not enough, we can enable charm debugging if that
> would help.
>
I'm not exactly sure at the moment what I'll need to get this race
condition to expose itself. Let me give it some thought and take a
closer look at the logs.
On 06/01/2016 09:41 AM, Andreas Hasenack wrote:
> We don't use deployer or amulet. Landscape talks to the juju api
> natively.
>
> We "reserve" machines beforehand via add-machine (both machines and
> containers) after bootstrap, and then we deploy services onto them. We
> also use the juju run trick twice to wait for relations to settle.
>
> There is no simple way to emulate landscape's deployment, you will have
> to use it yourself I guess, or figure out the ordering of events via the
> log files.
>
> If these logs are not enough, we can enable charm debugging if that
> would help.
>