On (15/11/16 10:03), Lukas Slebodnik wrote: >On (15/11/16 09:39), Jakub Hrozek wrote: >>On Wed, Nov 09, 2016 at 04:44:12PM +0100, Petr Cech wrote: >>> Hi all, >>> >>> I came back to ticket #3222 "sssd still showing ipa user after removed from >>> last group" [1]. And I have new knowledge. But I still do not see the light >>> at the end of the tunnel. >>> >>> [1] https://fedorahosted.org/sssd/ticket/3222 >>> >>> I attached patch which enables some basic debug on using of memcache. And >>> two reproducers (with and without memcache) which are based on reproducer >>> written in ticket. >>> >>> If we use memcache, the issue occurs only sometimes. >>> >>> The difference between both cases is mixed state of switch after >>> sss_nss_mc_getgrnam() call in _nss_sss_getgrnam_r() function. >>> >>> Note: code says (for default case): >>> /* if using the mmaped cache failed, >>> * fall back to socket based comms */ >>> >>> >>> Could anyone help, please? >> >>Hi Petr, >> >>this really seems like an issue with memory cache. I'm not an expert >>there, unfortunately, so I wonder what Michal and Lukas think.. >I cannot see any problem with memory cache here. >The same data are stored to memory cache as are returned >to client. If there is a bug then the bug is before storing >data to memory cache. They might be caused by some race condition caused >by different timeout for various entries in memory cache. > BTW, It looks like I can reproduce similar intermitent failures with downstream test (user-membership-list) as well.
The only problem is that I cannot reproduce it on el6.8 which has sssd-1.13.3-22. And memory cache was changed last time in sssd-1.13.2 (https://fedorahosted.org/sssd/ticket/2726) The bug has to be on responder side. LS _______________________________________________ sssd-devel mailing list -- sssd-devel@lists.fedorahosted.org To unsubscribe send an email to sssd-devel-le...@lists.fedorahosted.org
