On Fri, 2012-10-26 at 18:33 +0200, Michal Židek wrote: > On 10/24/2012 05:04 PM, Simo Sorce wrote: > > In sssd_nss you also need to take a F_WRLCK on the first byte of the > > file, however in sssd_nss case you want to retry a few times just in > > case you races with sss_cache. I think retrying a couple of times > > waiting a few milliseconds between each retry would be fine. > > > > Say 3 retries waiting 50ms between each. > > > > Ok. I used this for both old (to be unlinked) and new (created) memcache > files. But why is F_SETLKW not sufficient here (sssd_nss)? Are you > afraid of possible deadlock (it could happen if more sssd's were running > at the same time, like it happened recently because of a bug with > pidfile creation)? Or is there some other reason?
Yes and no. We can't wait forever because the monitor would kill us after a while if we do not respond to pings and then restart the sssd_nss process, which would probably get stuck again, but the admin wouldn't know why that is happening. It is better if the sssd_nss process fails fast with a clear critical error message (old level 0) in the logs. > NOTE: This patch applies on top of the recently posted patch in (see > thread: [SSSD] [PATCH] sss_cache: Multiple domains not handled properly) > > I tested it and it worked well for me. Any comments appreciated. Will comment on the patch soon. Simo. -- Simo Sorce * Red Hat, Inc * New York _______________________________________________ sssd-devel mailing list sssd-devel@lists.fedorahosted.org https://lists.fedorahosted.org/mailman/listinfo/sssd-devel