On Tue, 2012-12-04 at 14:59 -0500, Qing Chang wrote: > On 04/12/2012 2:50 PM, Jakub Hrozek wrote: > > On Tue, Dec 04, 2012 at 02:38:34PM -0500, Simo Sorce wrote: > >> On Tue, 2012-12-04 at 14:31 -0500, Qing Chang wrote: > >>> On 04/12/2012 2:19 PM, Simo Sorce wrote: > >>>> On Tue, 2012-12-04 at 13:11 -0500, Dmitri Pal wrote: > >>>>>> A "service sssd restart" "fix" the problem... > >>>>>> > >>>>>> I am thinking of running a cron job of above every 10 minutes to > >>>>> work > >>>>>> around > >>>>>> the problem while it is being troubleshoot. Is it a bad idea? > >>>> Well every 10 min. seem a bit too much, but I guess it depends how long > >>>> your user is willing to wait when it breaks :-/ > >>>>> This thread reminded me of the following thread > >>>>> https://lists.fedorahosted.org/pipermail/sssd-devel/2011-September/007058.html > >>>> Yes Dmitri, this is very similar. > >>>> > >>>> Qing, > >>>> do you happen to have the provider logs as well ? > >>>> > >>>> It seemed confined to the nss responder, but it may be an actual > >>>> provider issue after all, seem like the provider is deleting the user > >>>> from the cache from time to time, and it shouldn't. > >>>> > > The provider has no means of deleting a user from cache, in the current > > model, the responders are pure "readers" and only the back ends are > > "writers". > > > >>>> Simo. > >>> I am sorry, but what is the provider log? > >> Oh sorry, I should have been more specific, the provider log is the log > >> file named after your domain sssd_my.domain.name.log which contains logs > >> about the connection to the server and might explain what error the > >> provider got when it decided to go oflline (which may have also caused > >> the incorrect deletion of the user from the cache). > > I agree the issue seems to be on the provider side, though. For some > > reason, the provider is going offline and it shouldn't. > > > > What puzzles me is that Qing described the failing case as "open > > Thunderbird". I don't quite understand how that could fail except for a > > timeout? We should still return data from cache.. > sorry for the confusion, Jakub. I was just being specific that I asked her to > close > thunderbird and reopen it to trigger authentication attempt. At that point > dovecot2 already lost her. > > I have turned debug logging including provider log. Hopefully it happens > before > /var is fill up...
Feel free to turn down the NSS log if you want, or maybe set it to 0x0040 so you can more easily correlate the exact time when you get back this error: (Tue Dec 4 10:50:49 2012) [sssd[nss]] [nss_cmd_getpwnam_dp_callback] (0x0040): Unable to get information from Data Provider Simo. -- Simo Sorce * Red Hat, Inc * New York _______________________________________________ sssd-devel mailing list [email protected] https://lists.fedorahosted.org/mailman/listinfo/sssd-devel
