#1584 invalidating the memcache with sss_cache doesn't work if the sssd is not running
Closed: Fixed None Opened 7 years ago by jhrozek.

Stef found this corner case while preparing for the test day. If the SSSD is not running, for instance because the machine has left the domain, the memory cache files are still around and the processes that started before the machine left the domain can still query users from the domain.

The memcache can't even be invalidated with sss_cache, because sss_cache only sends a signal to the NSS responder.

This is how we should fix it:
1. make the API to invalidate the cache public
2. in the sss_cache tool, carefully detect if the sss_nss process is running, for example by checking for the responder socket or the PID.
3. if the sss_nss process is running, proceed with sending the signal
4. if the sss_nss process is not running, invalidate the memcache.

We should also check if all of the above works with SELinux.

Fields changed

milestone: NEEDS_TRIAGE => SSSD 1.9.3

Fields changed

owner: somebody => mzidek

Fields changed

design: =>
design_review: => 0
fedora_test_page: =>
patch: 0 => 1
status: new => assigned

- 33cbb78
- 0a55f90
- 304ce93
- b17eae0

resolution: => fixed
status: assigned => closed

Metadata Update from @jhrozek:
- Issue assigned to mzidek
- Issue set to the milestone: SSSD 1.9.3

2 years ago

Login to comment on this ticket.