On Monday 27 February 2006 18:32, Alan Brown wrote: > On Sun, 26 Feb 2006, Ian Levesque wrote: > >>> I've noticed a pattern in my backups when two jobs are simultaneously > >>> writing to the same tape. When the tape ends, Bacula doesn't respond > >>> gracefully. This is only a problem when two jobs are running on the > >>> tape, not when a single job is running. > >> > >> Are you spooling to disk or writing directly to tape? > > > > I'm writing directly to tape. Do you suppose that this may be remedied by > > spooling? My initial tests with spooling showed a performance hit > > On one job there's a performance hit _while writing_ > > On multiple jobs there's a performance hit _while reading_ because all the > jobs are interleaved, resulting in shoeshining. > > On multiple jobs, using slightly staggered start times, one job can be > spooling to disk while the other is dumping to tape (this will naturally > happen after a couple of spool/dump operatiopns anyway), givening an > overall speed increase, especially on incrementals, differentials > or backups of clients with sub-Gb/s connectivity > > Advantage 1: Backup jobs are clustered on the tape, making reading a far > cleaner streaming affair (read spool dump, skip to next section, no shoe > shining). > > Advantage 2: On incremental/differential jobs it pretty much eliminates > shoeshining while running the job. > > Advantage 3: Only one job is ever accessing the tape at a time, so the > issue you've hit won't happen. > > > It's clear you've tickled a bug and I hope you've filed it on > bugs.bacula.org.
I think it is more likely that the user did not properly test with btape test. Only one thread can write to the tape at any time, and when the end of the tape is hit, the tape remains in the exclusive use of the thread writing it. The return status that Bacula received was a -1 which should never be given by a modern tape drive at the end of a tape. I cannot completely exclude a bug, but I find it unlikely. I'd like clear evidence of a bug, which I don't have, before spending any time on the problem. > > However spooling will work around it and as soon as you're running > multiple simultaneous jobs you should consider using it anyway as it > usually gives _overall_ (not individual) speed advantages. > > > AB > > > > ------------------------------------------------------- > This SF.Net email is sponsored by xPML, a groundbreaking scripting language > that extends applications into web and mobile media. Attend the live > webcast and join the prime developer group breaking into this new coding > territory! > http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642 > _______________________________________________ > Bacula-users mailing list > Bacula-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/bacula-users ------------------------------------------------------- This SF.Net email is sponsored by xPML, a groundbreaking scripting language that extends applications into web and mobile media. Attend the live webcast and join the prime developer group breaking into this new coding territory! http://sel.as-us.falkag.net/sel?cmd=lnk&kid=110944&bid=241720&dat=121642 _______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
