On Fri, 2013-02-01 at 13:20 +0100, Thibault Le Meur wrote: > > On Fri, 2013-02-01 at 10:43 +0100, Thibault Le Meur wrote: > > Are you sure that duplicate suppression was enabled on the replica? > Yes I'm sure, and a simple find have quickly confrimed this. > find . -type f -a \! -links 1 -ls > .. > 141637005 32 -rw------- 12 cyrus mail 30678 sept. 20 > 18:42 ./obfustcated-username/12307. > ... Then something is deduplicating. > > I have "duplicatesuppression: yes" set on the replica. > I thought that duplicatesuppression was a different thing. Ah, yes, it is. We are talking about single-instance-store. That is the "singleinstancestore" directive. > What I'm trying to achieve is to keep the "Single Instance Store" > property on my replica > (http://cyrusimap.web.cmu.edu/docs/cyrus-imapd/2.4.7/overview.php#singleinstance) Yep. > In this case, since I use a single sync_client process per user, it is > logical that hardlinks between mailboxes can't be preserved. True, I guess that makes sense. I believe I did an initial migration using rsync and they fired up the sync-server to keep it up to date / get the last changes. But it was some time ago. It certainly dedupp'ed going forward. Perhaps sync-ing across different versions may effect it as well. > > I believe that skiplist is 32bit/64bit neutral. So convert to > > skiplist first, on the working server. KILL BERKELEY!! You want to > > do that anyway. > Yes it seems so. > When you say Kill Berkeley, do you mean that the DB should saty in > Skiplist format even on the production server (and not converted back > to Berkeley)? If yes, do you have any URL describing the procedure ? Yes, production 2.4.x boxes should be using Skiplist. The cvt_cyrusdb will convert databases between formats. ---- Cyrus Home Page: http://www.cyrusimap.org/ List Archives/Info: http://lists.andrew.cmu.edu/pipermail/info-cyrus/ To Unsubscribe: https://lists.andrew.cmu.edu/mailman/listinfo/info-cyrus