I was just looking at logs of a failed Fedora 27 test in openQA and noticed that ipa-dnskeysyncd was crashing constantly; that didn't seem related to the bug I was actually investigating, so I checked the logs from a *passed* test too, and sure enough, the same thing happens there. It seems that these crashes don't actually cause anything the tests do to fail, so we hadn't noticed it before.

The service keeps starting up, crashing, then waiting a while (due to systemd service holdoff times), then starting again and crashing again. It crashes like this:

Aug 17 10:14:49 ipa001.domain.local systemd[1]: ipa-dnskeysyncd.service: Service hold-off time over, scheduling restart.
Aug 17 10:14:49 ipa001.domain.local systemd[1]: Stopped IPA key daemon.
Aug 17 10:14:49 ipa001.domain.local audit[1]: SERVICE_START pid=1 uid=0 auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0 msg='unit=ipa-dnskeysyncd comm="systemd>
Aug 17 10:14:49 ipa001.domain.local audit[1]: SERVICE_STOP pid=1 uid=0 auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0 msg='unit=ipa-dnskeysyncd comm="systemd">
Aug 17 10:14:49 ipa001.domain.local systemd[1]: Started IPA key daemon.
Aug 17 10:14:49 ipa001.domain.local audit[1]: SERVICE_START pid=1 uid=0 auid=4294967295 ses=4294967295 subj=system_u:system_r:init_t:s0 msg='unit=ipa-dnskeysyncd comm="systemd>
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]: Traceback (most recent call last):
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:   File "/usr/libexec/ipa/ipa-dnskeysyncd", line 28, in <module>
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:     api.finalize()
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:   File "/usr/lib/python2.7/site-packages/ipalib/plugable.py", line 738, in finalize
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:     self.__do_if_not_done('load_plugins')
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:   File "/usr/lib/python2.7/site-packages/ipalib/plugable.py", line 425, in __do_if_not_done
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:     getattr(self, name)()
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:   File "/usr/lib/python2.7/site-packages/ipalib/plugable.py", line 619, in load_plugins
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:     self.add_package(package)
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:   File "/usr/lib/python2.7/site-packages/ipalib/plugable.py", line 646, in add_package
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:     module = importlib.import_module(name)
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:   File "/usr/lib64/python2.7/importlib/__init__.py", line 37, in import_module
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:     __import__(name)
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:   File "/usr/lib/python2.7/site-packages/ipaserver/plugins/certmap.py", line 22, in <module>
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]:     import dbus
Aug 17 10:14:50 ipa001.domain.local ipa-dnskeysyncd[8422]: ImportError: No module named dbus

which looks a lot like a missing dependency on dbus-python, probably in python2-ipaserver, which is where /usr/lib/python2.7/site-packages/ipaserver/plugins/certmap.py comes from.

This doesn't seem to be happening on F28 or newer. I think this may be because we never run this as Python 2 in F28+, and python3-ipaserver *does* have a dependency on python3-dbus. In F27 it seems both python3-ipaserver *and* python2-ipaserver get installed, and obviously there's still a Python 2 codepath that hits this bug.