Open pki-bot opened 4 years ago
Comment from edewata (@edewata) at 2016-07-28 00:09:14
Fixed in master:
Comment from edewata (@edewata) at 2016-09-07 01:29:49
The upgrade script has been removed temporarily due to a problem reported in ticket 2452.
Comment from mharmsen (@mharmsen) at 2016-09-12 22:42:47
Per CS/DS meeting of 09/12/2016:
Comment from cheimes (@tiran) at 2016-09-15 18:39:33
Does the restorecon command work properly during upgrade? We can run restorecon with the subprocess module:
subprocess.check_call(['/usr/bin/restorecon', '-R', '-F', '-v', instance.base_dir])
Comment from mharmsen (@mharmsen) at 2016-09-15 22:24:49
After a lengthy conversation on IRC on 09/15/2016, I have decided to move this ticket to 10.4.0 until such time as we can devise some sort of reproducer such that an SELinux bug may be filed.
Comment from edewata (@edewata) at 2016-09-16 05:58:37
Also discussed, another solution is to execute the upgrade scriptlets during PKI server restart instead of during RPM upgrade where presumably the SELinux tools will run successfully.
Comment from edewata (@edewata) at 2016-10-24 18:10:00
As noted in comment 10, this will be addressed in 10.4 by fixing the upgrade tool to run during server restart.
Comment from edewata (@edewata) at 2017-02-27 14:06:55
Metadata Update from @edewata:
This issue was migrated from Pagure Issue #2421. Originally filed by edewata (@edewata) on 2016-07-27 19:27:31:
Some files in the PKI instance folder were assigned incorrect SELinux contexts (e.g. unconfined_u instead of system_u), possibly because they were created after the SELinux contexts were set up.
For example:
A possible solution is to move the execution of the selinux_setup deployment scriptlet after all instance files are created, i.e. after security_databases scriptlet.
An upgrade script needs to be written to fix existing instances.