On Thu, Feb 2, 2012 at 3:33 PM, Thomas Nikolajsen <[email protected] > wrote:
> >commit 6de803f0514ff0019181d026a825d1e8a9a4c404 > >Author: François Tigeot <[email protected]> > >Date: Wed Jan 18 10:36:56 2012 +0100 > > > > Disable HAMMER live dedup, mark as experimental > > > > * Comment out mentions to it in the hammer(8) manpage > > > > * Live-dedup is a relatively new feature that has not undergone > > extensive testing. > > > > * It may have some issues with HAMMER bigblocks, we do not want > > users to run it on production machines yet. > >.. > >-SYSCTL_INT(_vfs_hammer, OID_AUTO, live_dedup, CTLFLAG_RW, > >- &hammer_live_dedup, 0, "Enable live dedup"); > >+SYSCTL_INT(_vfs_hammer, OID_AUTO, live_dedup, CTLFLAG_RD, > >+ &hammer_live_dedup, 0, "Enable live dedup (experimental)"); > > Any special reason to turn vfs.hammer.live_dedup sysctl to R/O? > (I haven't noticed a tunable to turn it on; only way is to change code) > > Do we have any known problem with live_dedup? > (I haven't noticed a bug on this in our bug handling system) > > Live_dedup was off by default; > which seemed fine for an experimental feature. > > -thomas > > It can be enabled from the loader / loader.conf. Live dedup was suspected of data corruption (but not confirmed) -- the thought was to make sure people really, really mean to turn it on for 3.0, it could be switched back to the way it was in master now that 3.0 has been branched. Sam
