Bug 1803134
| Summary: | Improve "unlock" time when user session already active | ||
|---|---|---|---|
| Product: | Red Hat Enterprise Linux 8 | Reporter: | hgraham |
| Component: | sssd | Assignee: | Sumit Bose <sbose> |
| Status: | CLOSED ERRATA | QA Contact: | sssd-qe <sssd-qe> |
| Severity: | high | Docs Contact: | |
| Priority: | unspecified | ||
| Version: | 8.1 | CC: | atikhono, dminnich, grajaiya, jhrozek, lslebodn, mniranja, mzidek, pbrezina, sbose, sgoveas, ssorce, stephenfin, thalman, tscherf, vvasilev |
| Target Milestone: | rc | Keywords: | Triaged |
| Target Release: | 8.0 | Flags: | pm-rhel:
mirror+
|
| Hardware: | Unspecified | ||
| OS: | Linux | ||
| Whiteboard: | sync-to-jira | ||
| Fixed In Version: | sssd-2.3.0-4.el8 | Doc Type: | If docs needed, set a value |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2020-11-04 02:04:37 UTC | Type: | Bug |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
| Bug Depends On: | |||
| Bug Blocks: | 1842946 | ||
|
Description
hgraham
2020-02-14 14:44:11 UTC
Henry (or Sumit), when following up on this bug, please fill also Description of the Problem, Severity and Reproduction scenario. This is needed for following reviews, testing, documentation, etc. Description of problem: Because of the initgroups call performed by sssd, users that have an active session on a system are seeing subsequent logins taking 10-20seconds. This affects usability and is hindering our roll out of IPA to CSB. Steps to Reproduce: 1. Set up a CSB RHEL8 server 2. Join to RHIT's IPA environment Actual results: unlocking a machine with an already active user session takes 10-20seconds Expected results: Subsequent unlocking only takes a few seconds Additional info: Upstream with patches that help in 8.0 but not 8.1 https://pagure.io/SSSD/sssd/pull-request/4104 Is there anything else needed from me for this? master: - 272efe4 - pam: make sure initgr cache is not created twice - 74f0a45 - cache_req: no refresh with CACHE_REQ_BYPASS_PROVIDER - 68aa68e - pam: use pam_initgroups_scheme - b66f0e4 - pam: add option pam_initgroups_scheme - d2424bf - pam: Use cache for users with existing session - b572871 - cache_req: introduce cache_behavior enumeration Upstream ticket: https://github.com/SSSD/sssd/issues/5062 Pushed PR: https://github.com/SSSD/sssd/pull/5222 * `master` * 100839b64390d7010bfa28552fd9381ef4366496 - PAM: do not treat error for cache-only lookups as fatal I just use brew download-build to get sssd-2.3.0-9.el8 9and all its dependencies) I see no appreciable difference in behavior after the update. I ran a sudo -i command and it took it 10 seconds to give me a password prompt, then it took another 10 second to complete authentication. This is consistent with a previous report that my original patchset stopped working sometimes between around 8.1+ as some other code is now forcing a synchronous LDAP call in the code path. The way to test this is the following: configure SSSD against an IPA server. - log in a user. - cause the connection to the server to fail (restart ldap on the server for example). - try to use sudo or other call that causes authentication. - *verify* that no attempt to make LDAP calls happen until after the password prompt is shown HTH (IMO this is a FailQA situation) Henry, for which platform do you want a build? Simo, can you attach your current logs? bye, Sumit Hi Sumit, that is exactly the issue. I searched the timestmaps db and it was empty except for the basic skeleton tree entries. After adding the entry and refreshing the user now sudo access is very quick, sub-second delays. The same also for GDM logins, no more delays to unlock the screen. Thanks a lot, seem that fixing that bug is quite important for a decent experience. My hypothesys is that a database migration over an upgrade failed to port/create proper timestamps. The code should be fixed to cretae a timestamp on update if an entry does not exist I guess. In any case, thanks a lot for diagnosing this. It means the unlock patches work as expected now and all is good from my side!! (In reply to Simo Sorce from comment #32) > Hi Sumit, > that is exactly the issue. > > I searched the timestmaps db and it was empty except for the basic skeleton > tree entries. > > After adding the entry and refreshing the user now sudo access is very > quick, sub-second delays. > The same also for GDM logins, no more delays to unlock the screen. Great, so this explains as well, why I didn't see such strong delays and unexpected initgroup requests in my test. > > Thanks a lot, seem that fixing that bug is quite important for a decent > experience. > > My hypothesys is that a database migration over an upgrade failed to > port/create proper timestamps. The code should be fixed to cretae a > timestamp on update if an entry does not exist I guess. Yes. > > In any case, thanks a lot for diagnosing this. > It means the unlock patches work as expected now and all is good from my > side!! yw, thanks for your patience. bye, Sumit *** Bug 1795127 has been marked as a duplicate of this bug. *** Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory (sssd bug fix and enhancement update), and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHBA-2020:4569 |