maas-devel team mailing list archive

[Martin Packman <martin.packman@xxxxxxxxxxxxx>]

On 10/10/2012, Diogo Matsubara <matsubara@xxxxxxxxxxxxx> wrote:
> Hi there,
>
> today I tested maas - 0.1+bzr1239+dfsg-0+1240+125~ppa0~quantal1 from
> the daily builds PPA.

Thanks for doing all this!

>   - memory set to 0 in the lenovo nodes:
> https://bugs.launchpad.net/maas/+bug/1064638

This involves the lenovo node giving very different lshw output to
what we've seen so far. It's trivial enough to fix the expression we
use to lookup memory, but it makes me worry a little about how stable
this will be longer term.

>   - tag commands don't have useful help:
> https://bugs.launchpad.net/maas/+bug/1064735

John has a branch up fixing some of the docstrings here.

>   - passing empty value in the tag definition argument seems to hang
> forever: https://bugs.launchpad.net/maas/+bug/1064737

Jeroen has tracked this down to probably being a bad mod_wsgi/lxml
interaction, there's a workaround that involves telling mod_wsgi to
only run the code in the initial sub-interpreter that should tell us
if that's the issue here.

>   - couldn't use tags to deploy with juju. Got an error "no such tag":
> https://bugs.launchpad.net/maas/+bug/1064734

This is painful as juju 0.6 was released yesterday, but we need the
fix there. Short story, my initial code version involved a set and
used ",".join when passing on to api, then simplified to just pass a
string straight through, then in response to review re-added the set
logic to pre-validate tags but missed adapting the api back again. So
MaaS is passed the stringified set rather than the original tag
string. The provider needs some better unit test coverage for
constraint/api interactions.

Martin