Skip to content

Service.is_valid always returns invalid because of unrelated issues #689

Open
@coofercat

Description

@coofercat

On a system with some systemd issues, the outcome of host.service('myservice').is_valid is always False, even if my service is actually good.

On further investigation the original pull request to add this #265 added a check to make sure STDOUT and STDERR were emtpy, as the call to systemd-analyze verify %s always returns 0. However, in come cases, STDERR can contain unrelated information which trips up the comparison.

As an example, I have a deliberate typo in my service unit. The output of systemd-analyze verify myservice.service is as follows (all on STDERR):

snap-core18-2632.mount: Unit is bound to inactive unit dev-loop1.device. Stopping, too.
snap-snapd-18357.mount: Unit is bound to inactive unit dev-loop5.device. Stopping, too.
snap-snapd-17883.mount: Unit is bound to inactive unit dev-loop4.device. Stopping, too.
snap-core20-1828.mount: Unit is bound to inactive unit dev-loop7.device. Stopping, too.
/lib/systemd/system/myservice.service:17: Missing '=', ignoring line.

If I correct the typo, then the last line of this output goes away, but the others remain (actually, on my system the others seem to change every time I run the verify, but I always get at least a couple of lines out of it).

I'm not sure what's up with my system that it should do this (or why systemd would behave this way) but I'm wondering what the best solution to the testinfra problem is. I'm thinking a sort of "grep" of the output for myservice.service would suffice? Something like:

stderr = "".join(filter(lambda x: name in x, cmd.stderr.split()))

Or in place in the function in testinfra/modules/service.py (also taking the opportunity to add some output to explain the reason for the failure):

   @property
    def is_valid(self):
        # systemd-analyze requires a full path.
        if self.name.endswith(".service"):
            name = self.name
        else:
            name = self.name + ".service"
        cmd = self.run("systemd-analyze verify %s", name)
        # A bad unit file still returns a rc of 0, so check the
        # stdout for anything.  Nothing means no warns/errors.
        # Docs at https://www.freedesktop.org/software/systemd/man/systemd
        # -analyze.html#Examples%20for%20verify
        stderr = "".join(filter(lambda x: name in x, cmd.stderr.split()))
        assert (cmd.stdout, stderr) == ("", ""), "service %s is not valid (see 'systemd-analyze verify %s' for details)" % (name, name)
        return True

Happy to hear other suggestions or ideas etc :-)

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions