Hmmm - I don't get that: > you can not discard/recycle the differential before a new full
The way I understand it is that differential is everything changed since the last full (unlike some other systems definitions). That to me means that after a new differential, you can abandon all older incremental and differential backups since the last full, and start a new incremental cycle from here, unless of course you want to keep the ability to go back in time, in which case you can also chose to keep those. this gives as far as I can see exactly what you want: > so when you need to restore you just need to pull the > incremental back to the most recent differently and the full. You should > be able to then discard differentials at any time as a space/restore speed > tradeoff. Are there some obvious or subtle points that I don't get here? Concerning the other points - I totally agree about the 'timestamp / not including all files point and I simply don't understand why that didn't come up as no 1 in the vote for features - but it didn't. Same goes for the 'deleted files point'. What I don't understand in relation to these points is that it is often mentioned that implementing them would be resourcedemanding in terms of comparing every file with the ones in the catalog. I don't understand why it can't be done the other way around, so that youjust take a kind of snapshot of the filesystem and store that, so that you only use it in case of a restore - then you can use it to delete deleted files and move the moved ones. Concerning choice of DB - I think it is underlined in the manual that the built in DB is only for testing or small scale setups. Rpm's that I used on Mandriva worked out of the box - except for the configurations of course, which are quite extensive to dig into and get to know. On the other hand that is what makes it so flexible, and once you get to know it it is very quick to make small changes or additions to the system. My additions to the pros: Bacula comes with some very nice prebuilt commands and queries that immediately gives me the information I need both about the state of the system and the history of backups and media. These I had to build myself in Networker, and to figure that out took me longer than my whole bacula configuration. Backup to disk seems to work very well - and working configuration example have been given in the manual. Monday 24 April 2006 20:54, Eric Warnke wrote: > Clarification on the differentials comment > > What I'm suggesting is... > > Full = Full > Incremental = Changes since last Full or Incremental > Differential = Changes since last Full ( effectively becoming an > incremental since last full ). > > Right now, incremental track from either full, differential, or incremental > meaning you can not discard/recycle the differential before a new full, > what I'm suggesting is that should not be the case. By decoupling the > incremental from differentials you can manage very long retention periods ( > ie in our case 90+ days ). Differentials would then be a "roll-up" of the > incremental so when you need to restore you just need to pull the > incremental back to the most recent differently and the full. You should > be able to then discard differentials at any time as a space/restore speed > tradeoff. > > Cheers, > Eric > > On 4/24/06, Meidal, Knut <[EMAIL PROTECTED]> wrote: > > 4) The way that incremental and differentials are structured does not > > allow > > for good differential pooling since incremental can be based on > > differentials. Because of this you can't recycle your differentials > > before > > you re-run a full backup. If incremental were only based on fulls or > > incremental, you can recycle your differentials on a faster cycle to > > reduce > > tape space for in exchange for longer restore times for older files. > > > > -> That is an interesting point. Any subsequent differentials obsoletes > > previous when aiming for recovering to most current state. > > The problem I see, is that recycling a differential will orphan all the > > incrementals referring to them(between diff and new diff or full), making > > them useless. > > I can't seem to understand your second point if incrementals were only > > based > > on full or inc... > > If you wanted to that, you could simply remove the diffs from the > > schedule, > > and never run one? I can't seem to see the useability of a lone > > differential > > that can't be used for baseline for subsequent incrementals. > > I've been known to misunderstand, so bear with me.. > > > > > > > > Knut Meidal -- Regards Steen ------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
