On Thu, Dec 16, 2021 at 10:47 PM Nir Soffer <nsof...@redhat.com> wrote:
> On Thu, Dec 16, 2021 at 10:14 PM Milan Zamazal <mzama...@redhat.com>
> wrote:
> >
> > Hi,
> >
> > when I run Vdsm tests as a non-root user in vdsm-test-centos-8
> > container, they systematically fail on several storage tests. I run the
> > tests as
> >
> > podman run -it -v $HOME:$HOME --userns=keep-id vdsm-test-centos-8
> $HOME/test.sh
>
> Why do you need the --userns option?
>
> You run this in a container because your development environment
> is not compatible - right?
>
> > where test.sh is
> >
> > #!/bin/sh
> >
> > export TRAVIS_CI=1
>
This is not correct for your local container - in travis we run a
privileged container
as root and we create loop devices before the test.
> > cd .../vdsm
> > ./autogen.sh --system
> > make clean
> > make
>
> I never run autogen and make during the tests, this can be done once
> after checkout or when modifying the makefiles.
>
> > make lint
> > make tests
>
> You are missing:
>
> make storage
>
> Without this, a lot of tests will be skipped or xfailed.
>
> > make tests-storage
> >
> > The failing tests are in devicemapper_test.py, outofprocess_test.py and
> > qemuimg_test.py. I have also seen a test failure in nbd_test.py but not
> > always. Is it a problem of the tests or of my environment?
>
> nbd_test.py should be skipped unless you run as root, or have running
> supervdsm serving your user.
>
> I usually run the storage tests like this locally on Fedora (35 now):
>
> make storage
> tox -e storage
>
> Results:
>
> storage/devicemapper_test.py - 5 pass, rest skip
> storage/nbd_test.py - 2 pass, rest skip
> storage/qemuimg_test.py - 2 skips
> storage/outofprocess_test.py - 3 skips, rest pass
>
> 2350 passed, 161 skipped, 102 deselected, 3 xfailed, 13093 warnings in
> 134.80s (0:02:14)
>
> I'll try later using the container.
>
>
I tried this:
podman run --rm -it -v `pwd`:/src:Z --userns=keep-id vdsm-test-centos-8
cd src
make tests-storage
Got 2 test failures:
============================================================= FAILURES
==============================================================
______________________________________________________
test_block_device_name
_______________________________________________________
def test_block_device_name():
devs = glob.glob("/sys/block/*/dev")
dev_name = os.path.basename(os.path.dirname(devs[0]))
with open(devs[0], 'r') as f:
major_minor = f.readline().rstrip()
> assert devicemapper.device_name(major_minor) == dev_name
E AssertionError: assert '7:1' == 'loop1'
E - loop1
E + 7:1
This likey failed since there are no loop devices in the container:
bash-4.4$ ls /dev/
console core fd full mqueue null ptmx pts random shm stderr
stdin stdout tty urandom zero
And there is no way to create them, since your run as regular user
and sudo does not work. Event if it works I think you will not be able
to create the loop devices since the container is not privileged.
It may be possible to map the loop devices from the host to the container
but I never tried.
storage/devicemapper_test.py:243: AssertionError
___________________________________________________
test_stop_server_not_running
____________________________________________________
@broken_on_ci
def test_stop_server_not_running():
# Stopping non-existing server should succeed.
> nbd.stop_server("no-such-server-uuid")
storage/nbd_test.py:806:
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
../lib/vdsm/storage/nbd.py:179: in stop_server
info = systemctl.show(service, properties=("LoadState",))
../lib/vdsm/common/systemctl.py:74: in show
out = commands.run(cmd).decode("utf-8")
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
args = ['/usr/bin/systemctl', 'show', '--property=LoadState',
'vdsm-nbd-no-such-server-uuid.service'], input = None, cwd = None
env = None, sudo = False, setsid = False, nice = None, ioclass = None,
ioclassdata = None, reset_cpu_affinity = True
def run(args, input=None, cwd=None, env=None, sudo=False, setsid=False,
nice=None, ioclass=None, ioclassdata=None,
reset_cpu_affinity=True):
"""
Starts a command communicate with it, and wait until the command
terminates. Ensures that the command is killed if an unexpected
error is
raised.
args are logged when command starts, and are included in the
exception if a
command has failed. If args contain sensitive information that
should not
be logged, such as passwords, they must be wrapped with
ProtectedPassword.
The child process stdout and stderr are always buffered. If you have
special needs, such as running the command without buffering
stdout, or
create a pipeline of several commands, use the lower level start()
function.
Arguments:
args (list): Command arguments
input (bytes): Data to send to the command via stdin.
cwd (str): working directory for the child process
env (dict): environment of the new child process
sudo (bool): if set to True, run the command via sudo
nice (int): if not None, run the command via nice command with
the
specified nice value
ioclass (int): if not None, run the command with the ionice
command
using specified ioclass value.
ioclassdata (int): if ioclass is set, the scheduling class
data. 0-7
are valid data (priority levels).
reset_cpu_affinity (bool): Run the command via the taskset
command,
allowing the child process to run on all cpus (default
True).
Returns:
The command output (bytes)
Raises:
OSError if the command could not start.
cmdutils.Error if the command terminated with a non-zero exit
code.
utils.TerminatingFailure if command could not be terminated.
"""
p = start(args,
stdin=subprocess.PIPE if input else None,
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
cwd=cwd,
env=env,
sudo=sudo,
setsid=setsid,
nice=nice,
ioclass=ioclass,
ioclassdata=ioclassdata,
reset_cpu_affinity=reset_cpu_affinity)
with terminating(p):
out, err = p.communicate(input)
log.debug(cmdutils.retcode_log_line(p.returncode, err))
if p.returncode != 0:
> raise cmdutils.Error(args, p.returncode, out, err)
E vdsm.common.cmdutils.Error: Command ['/usr/bin/systemctl',
'show', '--property=LoadState', 'vdsm-nbd-no-such-server-uuid.service']
failed with rc=1 out=b'' err=b"System has not been booted with systemd as
init system (PID 1). Can't operate.\nFailed to connect to bus: Host is
down\n"
This fails because we have systemd in the container, but we did not start
the
container in the write way to make it happy.
I'm not sure why the test was not skipped, probably a bug in the skip
condition.
2 failed, 1965 passed, 164 skipped, 102 deselected, 383 xfailed, 1 warning
in 150.30s (0:02:30)
I'll try to fix the 2 failing tests so they skip in this condition.
The first test can probably be skipped we don't have any /dev/loop*
The second can be skipped if systemctl show does not work.
Nir
_______________________________________________
Devel mailing list -- devel@ovirt.org
To unsubscribe send an email to devel-le...@ovirt.org
Privacy Statement: https://www.ovirt.org/privacy-policy.html
oVirt Code of Conduct:
https://www.ovirt.org/community/about/community-guidelines/
List Archives:
https://lists.ovirt.org/archives/list/devel@ovirt.org/message/GX446KG4GFWS5VYACG5AK6KNJQMF665O/
