On Sat, Jun 13, 2009 at 09:58:17PM +0100, Ed W wrote: > John A. Sullivan III wrote: > >Thanks! That saved me some serious grief - John > It would still be very interesting to try and find what's causing the > issue though. The logic of vhashify should only create hard links > between identical files. In fact I'm curious how/why it's even updating > directory times? because in unix creating or removing a file/link is a directory operation? > I don't have the logic of the hash open right now, but presumably we > need to unlink one side and hard link the otherside (or perhaps it > copies/moves one side into the cache folder and hard links everything > else?) The point being I guess if you had an open file which got > hardlinked then you might be unlucky enough to end up writing to > something which was now unlinked? that is entirely possible, but quite unlikey, because chances are small that an open database is identical to another > That your writes to the innodb log files have stopped suggests that > perhaps part of the log files got linked while open? However, this > seems unlikely given that the data files ought to be completely > different in all your vservers? correct, my guess would be that innodb is being too smart and deducates something from the fact that some time stamps for updated best, Herbert > Do you do anything special which might cause these files to look > identical before running vhashify? > Ed W