Red Hat Bugzilla – Bug 1005316
RHEL6.5 ipa-server-install --uninstall does not remove /var/lib/sss/pubconf/kdcinfo.$REALM
Last modified: 2013-09-12 07:41:57 EDT
Description of problem:
It appears that IPA uninstall is again not removing /var/lib/sss/pubconf/kdcinfo.$REALM
This was seen after an upgrade from RHEL6.2 to RHEL6.5:
:: [ FAIL ] :: Make sure that uninstall removed /var/lib/sss/pubconf/kdcinfo.TESTRELM.COM. Bug BZ 829070 (Expected 2, got 0)
That test simply does an ls check:
rlRun "ls /var/lib/sss/pubconf/kdcinfo.$RELM" 2 "Make sure that uninstall removed /var/lib/sss/pubconf/kdcinfo.$RELM. Bug BZ 829070"
Version-Release number of selected component (if applicable):
always (at least for upgrades).
Steps to Reproduce:
1. Install IPA (on rhel6.2)
2. Upgrade to RHEL 6.5 3.0.0 version of ipa
3. ipa-server-install --uninstall -U
4. ls -ld ls /var/lib/sss/pubconf/kdcinfo.$RELM
directory still exists
directory removed by uninstall
This is the same thing previously fixed for:
It's interesting to note that we didn't actually make any changes in https://bugzilla.redhat.com/show_bug.cgi?id=829070 , so I'm not sure we can call this a regression.
We have no cleanup code associated with that file. I'm not sure who is responsible for deleting it, though sssd owns it. I don't know if authconfig or sssd is supposed to handle it.
Interesting. I should have re-read through the whole bug there.
I may have been thinking of another bug where the IPA uninstall was going to be updated to handle removing something. Let me check an upgrade again.
Let me re-check something and run some more tests on this one. I'm removing the regression keyword for now as well.
I'll get back to you guys later today on this one.
Ok, I saw this again in automated tests but, I'm not seeing it on some of my manual tests. So, seems to be inconsistent.
I may need help next week tracking this down.
Jakub, can you please advise? As Rob said, IPA does not really touch the kdcinfo...
kdcinfo should be removed by the SSSD as soon as the provider goes offline or as soon as the SSSD is stopped. Does ipa-server-install --uninstall stop the SSSD?
If so, and you still see the file around, then it's and SSSD bug.
(In reply to Jakub Hrozek from comment #7)
> kdcinfo should be removed by the SSSD as soon as the provider goes offline
> or as soon as the SSSD is stopped. Does ipa-server-install --uninstall stop
> the SSSD?
> If so, and you still see the file around, then it's and SSSD bug.
Moving to SSSD for now, until Scott confirms if SSSD was properly stopped when the kdcinfo was still around.
I am thinking that this may be a race condition in the test itself which does not wait until SSSD is stopped and tests it before that.
So, ipa-server-install --uninstall doesn't wait for sssd to come down completely?
If so, I think Martin may be right about race condition there. Looking at the test output again, sssd is not yet shut down:
:: [ FAIL ] :: Make sure that uninstall removed /var/lib/sss/pubconf
/kdcinfo.TESTRELM.COM. Bug BZ 829070 (Expected 2, got 0)
root 15772 1 0 19:23 ? 00:00:01 /usr/sbin/sssd -f -D
root 15773 15772 0 19:23 ? 00:00:01 /usr/libexec/sssd/sssd_be --domain testrelm.com --debug-to-files
root 15774 15772 0 19:23 ? 00:00:00 /usr/libexec/sssd/sssd_nss --debug-to-files
root 15775 15772 0 19:23 ? 00:00:00 /usr/libexec/sssd/sssd_pam --debug-to-files
root 15776 15772 0 19:23 ? 00:00:00 /usr/libexec/sssd/sssd_ssh --debug-to-files
root 15777 15772 0 19:23 ? 00:00:00 /usr/libexec/sssd/sssd_pac --debug-to-files
:: [ FAIL ] :: Make sure that sssd appears to be stopped as per BZ 830598 (Expected 1, got 0)
So, That test for 820598 may simply be that the ps ran before sssd could stop?
Does the SSSD shuts down eventually? Like several seconds after uninstallation finishes.
It would be also useful to attach ipaserver-uninstall.log to find out if the sssd was really requested to stop and what was the result.
Also, if you have the SSSD logs handy, there should be be a log message telling you that the files are being removed.
I believe this is right...I haven't had a chance to get logs but, I added a 60 second sleep before the check and it's not failing anymore...
So, I believe it's a test problem as mentioned. I'm going to run a few more tests today and if I don't see anything else, I'll just close this one as NOTABUG.
I ran a local test as well and saw the files being removed correctly. Closing on my end.
Scott, kindly reopen if you still see the problem.