On 03/26/2012 04:44 AM, MORITA Kazutaka wrote: > What does it mean? bdrv_co_flush_to_disk needs the error value, > doesn't it?
I was thought to relax the flush operation a bit because, when the flush operation happens at a recovery stage, it might time-out fails (the targeted node down) and if we relax it, the next flush operation will flush the dirty bits. If you think this compromise is not good (trade off with consistency), I am okay to return err back. For practical usage, I think those that demand strong consistency should be suggested to not use object cache. Thanks, Yuan -- sheepdog mailing list [email protected] http://lists.wpkg.org/mailman/listinfo/sheepdog
