Open Stack

On Tue, 2019-04-23 at 16:50 -0500, Matt Riedemann wrote:
> On 4/23/2019 2:04 PM, Sean Mooney wrote:
> > > 2019-04-23 17:59:37.663 3926*ERROR*  nova.scheduler.utils
> > > [req-abd32ef2-3a19-4154-a7da-893f3d2b2aef a16c6ef0319b4643a4ec8e56a1d025cb
> > > 59065d8f970b467aa94ef7b35f1edab5 - default default] [instance:
> > > 269d2480-4477-4382-b09c-9c686ab7a17c] Error from last host: compute (node
> > > compute.cujae.edu.cu): [u'Traceback (most recent call last):\n', u'  File
> > > "/usr/lib/python2.7/dist-packages/nova/compute/manager.py", line 1841, in
> > > _do_build_and_run_instance\n    filter_properties, request_spec)\n', u'
> > > File "/usr/lib/python2.7/dist-packages/nova/compute/manager.py", line 2121,
> > > in _build_and_run_instance\n    instance_uuid=instance.uuid,
> > > reason=six.text_type(e))\n', u'RescheduledException: Build of instance
> > > 269d2480-4477-4382-b09c-9c686ab7a17c was re-scheduled: mount() takes
> > > exactly 4 arguments (3 given)\n']
> > 
> > the libvirt lxc dirver is not tested upstream in the gate so this could
> > be a latent bug. its hard to tell but it might be this mount function
> > https://github.com/openstack/nova/blob/c6218428e9b29a2c52808ec7d27b4b21aadc0299/nova/privsep/fs.py#L32-L40
> > 
> > being called here
> > 
https://github.com/openstack/nova/blob/c6218428e9b29a2c52808ec7d27b4b21aadc0299/nova/virt/disk/mount/api.py#L250-L251
> > where only 3 args are passed instead of the 4 that are needed
> > the options arg is missing.
> > but if thats the issue its been broken for years.
> > sincehttps://github.com/openstack/nova/commit/7ad72b092201f87530674a245e2904c6401d567c
> > 
> 
> Yup it's a latent bug. Feel free to report a bug and post a patch, but 
> you're likely building your house on sand because as noted the 
> libvirt+lxc driver doesn't get integration testing.
i think the lxc backend was original a focus of canonical but they pivoted to nova lxd instead.
if you want lxc container using nova-lxd it probably better tested but its also not in the gate.
> I used to try to 
> make a CI job work for that configuration but it was extremely flaky and 
> it was never sorted out what the underlying issues were (and no one 
> cared enough to take it up).
i can proably try to fix this issue and if i figure out how to deploy and test it with devstack i might be able
to get a gate job working but is this something peopel want fixed? from what i can tell this looks like its been
busted for 2 years so it seams like this does not get used by anyone who is active upstream at least.

the fix iteself whoudl eb relitivly simple but the fact this was not caguht by any of your unit tests indicate we
likely dont have any for this code or it not asserting the mocks are called correctly.

i might try following https://docs.openstack.org/devstack/latest/guides/lxc.html but that likely is jsut as well
maintined and tested...
>