On Thu, Aug 30, 2012 at 3:44 PM, Noah Watkins <[email protected]> wrote: > On Thu, Aug 30, 2012 at 3:39 PM, Sage Weil <[email protected]> wrote: >> On Thu, 30 Aug 2012, Noah Watkins wrote: >> >> I think this is just the mds reconnect delay; it's complete normal when >> the mds restarts and some of the clients also crashed that it has to wait >> for the full reconnect window to pass. > > Are you talking about the clients running during the restart? In the > case I'm describing I bring down all daemons and the client, before > starting the cluster and re-running the test.
How long a hang have you observed? Since you brought down a client, the MDS needs to wait through a full time-out period before letting other people do things (it's giving the previous client a chance to claim some state). The default looks to be 45 seconds right now. But the patch shouldn't have changed that behavior in any way; so if it's actually different behavior in testing then we just broke...something else, somehow... -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
