[00:19:14] <Simon Wilkinson> Actually, that would do it, I think. We flush, which actually pushes all of the pages to disk, but that races with pdflush, which sees a dirty entry in the radix tree, and writes the page again, after the close has completed. We're not seeing data corruption reports because the data has actually gone to disk with the first flush - the second is, in effect, a no-op.
[00:41:09] --- Russ has left: Disconnected
[02:04:32] --- Simon Wilkinson has left
[06:24:10] --- jaltman has left: Disconnected
[09:32:37] --- deason has become available
[12:03:19] --- jaltman has become available
[12:03:26] --- Russ has become available
[12:24:27] --- jaltman has left: Disconnected
[15:49:44] --- mdionne has become available
[17:48:53] --- mdionne has left
[20:12:33] --- deason has left
[20:50:09] --- asedeno has left
[20:50:21] --- asedeno has become available
[21:29:52] --- stevenjenkins has left