I run into this issue with several of my servers and dealt with it by creating "migrate" jobs. First job goes to disk. Second job runs some reasonable time later and migrates the D2D job to tape. I had a number of key servers I did this for with the advantage that I could offsite the tapes and keep the D2D job on disk till after the next backup had run. This way I had immediate recovery available as well as disaster recovery.
On Sun, Oct 26, 2014 at 7:59 AM, Harald Schmalzbauer < h.schmalzba...@omnilan.de> wrote: > Hello, > > I enable data spooling for almost any job, because my LTO4 drive's > hw-compression allows to stream _my_ data at little over 100MByte/s > average, which bacula-fd can't deliver (localhost FD-SD connections > allow ~25MB/s with 60+% CPU usage; SoftCompression is disabled; oberved > FD reqests are 4KB/t only, but that's a completely different issue I > have to look at more closely some times later). > I'm aware that the numbers above are highly workload dependent and are > not representative. > They basically should give an idea why I need data spooling – nothing > more. Environemnt is bacula 7.0.5 (dir, sd and fd), FreeBSD 10.1 (amd64) > (also dir, sd and fd), LACP-GbE (irrelevant since I'm using localhost > sockets), 3.4GHz Xeon-E3v3, 8GB RAM and two LSI2008. > > A typical job is 100-500GB in size. So spooling with only 25MB/s takes > significant ammount of time. > That's why I want to start the job as early as possible.. > My problem is that data spooling starts _after_ the volume has been > mounted and positioned. > So if I start the job at midnight, no single bit get's spooled unless > somebody feeds the correct tape next morning :-( And next morning very > often delays to "next lunch"; that's were the missing 4 hours breaks the > timetable for my setup. > > Is there already an option I missed, which would enable data spooling > without requesting the volume first? > That would save exactly the hours my setup needs as buffer if somebody > forgets to feed the tape. > Or is there any reason why data spooling must be delayed? I can't > imagine one. If something goes wrong at despooling, nothing changes if > volume was available before spooling or not. > > Thanks, > > -Harry > > > > > > ------------------------------------------------------------------------------ > > _______________________________________________ > Bacula-users mailing list > Bacula-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/bacula-users > > -- ------------------------------------------------------------------- John M. Lockard | U of Michigan - School of Information Unix Sys Admin | 105 South State St. | 4325 North Quad jlock...@umich.edu | Ann Arbor, MI 48109-1285 www.umich.edu/~jlockard <http://www.umich.edu/%7Ejlockard> | 734-936-7255 | 734-764-2475 FAX ------------------------------------------------------------------- - The University of Michigan will never ask you for your password -
------------------------------------------------------------------------------
_______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
