On (23/02/16 13:37), Jakub Hrozek wrote:
On Thu, Feb 18, 2016 at 02:04:54PM +0100, Lukas Slebodnik wrote:
> ehlo,
>
> It took me some time to reproduce issue with cron.
> It occured very rarely in my case (twice in a week).
>
> Therefore I prepared different reproducer "mini_cron.c"
> attached in mail. It tries to query for data in the interval
> (9.990 sec .. 10.069 sec) when responder might destroy connection.
> mini_cron expect there is a user mof_user1 in LDAP and client_idle_timeout
> has minimal value 10 seconds. Default is 60 seconds. It's also good to
> decrease memory cache timeout to ensure connection to responder every time.
>
> e.g.
> [sssd]
> config_file_version = 2
> services = nss, pam
> domains = LDAP
> client_idle_timeout = 10
>
> [nss]
> filter_groups = root
> filter_users = root
> memcache_timeout = 0
> client_idle_timeout = 10
> debug_level =9
> debug_microseconds = true
>
> Detailed explanation is in commit message.
>
> Attached is also a debug patch which I used as part of
> analysis when it can fail.
Thank you.
>
> BTW I ran mini_cron reprodurer for a week and it didn't fail.
>
> LS
The code looks good to me and sanity testing of the clients passed as
well -> ACK
CI is still running.
I realized that I didn't fix it in pam client.
LS