On 08/26/2012 06:50 PM, MORITA Kazutaka wrote: > This is necessary to pass test 034. > > Storing stale objects in a flat directory causes a problem. It is > because if the objects are updated while leaving cluster, sheep cannot > detect it and uses the local stale objects as the latest one. It is > mandatory to store stale objects with epoch information like old farm > code or simple store. > > This patch create a epoch directory like simple store. I think this > is the simplest way to store stale objects with epoch info.
I am actually doing the same problem. I have thought of this approach, but I come up with another one, instead of use [epoch]/oid to tag the stale object, I just play with its name to store stale objects: .stale/oid.epoch My patch is finished but still can't pass the updated 028. The problem your patch set don't handle is: for default_read(), when iocb.epoch < sys_epoch() we need read upwards like Farm did to fix subtle problems found in multiple node change events. I'm going to test your current devel branch against updated 028. Thanks, Yuan -- sheepdog mailing list [email protected] http://lists.wpkg.org/mailman/listinfo/sheepdog
