Empty log file when a container is started with the API

[Jean-Baptiste Lallement]

When a container is started with the python API the log file /var/log/lxc/NAME.log is empty. This is pretty annoying to debug pre-start hook issues for example.

ProblemType: Bug
DistroRelease: Ubuntu 13.10
Package: lxc 0.9.0-0ubuntu7
ProcVersionSignature: Ubuntu 3.9.0-1.5-generic 3.9.1
Uname: Linux 3.9.0-1-generic x86_64
ApportVersion: 2.10.1-0ubuntu1
Architecture: amd64
Date: Fri May 17 10:31:44 2013
MarkForUpload: True
SourcePackage: lxc
UpgradeStatus: Upgraded to saucy on 2012-01-31 (471 days ago)
lxcsyslog:

[Serge Hallyn (serge-hallyn)]

Thanks for reporting this bug.

If you do

c.set_config_item("lxc.log.loglevel', '1')

before starting the container, does that help?

Also, if you cause an actual error during startup (i.e. specify a wrong rootfs location in the configuration), does an error then at least show up in the logfile? If not, then that should be fixed.

[Didier Roche-Tolomelli (didrocks)]

I tried multiple things:
c.set_config_item("lxc.loglevel", "1")
c.set_config_item("lxc.loglevel", "9")
c.set_config_item("lxc.loglevel", "debug")

And I always get:
lxc: invalid log priority 9

Without those, setting in the lxc config:
lxc.logfile = /tmp/foo
lxc.loglevel = debug

it creates an empty /tmp/foo file, but no content at all, even if it errored out.

[Serge Hallyn (serge-hallyn)]

Thanks, this is likely related to how the python bindings are
interpreting the second command. I woudl actually expect
c.set_config_item('lxc.loglevel', 'info') to work, but that
fails here too.

 status: confirmed
 assignee: stgraber
 priority: medium