On Thu, Sep 14, 2006 at 08:34:41PM +0200, Kern Sibbald wrote: > On Thursday 14 September 2006 19:41, Jo Rhett wrote: > > On Thu, Sep 14, 2006 at 10:20:21AM +0200, Kern Sibbald wrote: > > > On Thursday 14 September 2006 03:45, Jo Rhett wrote: > > > > > > So right now the "Volume Use" start time is set by the start time > > > > > > of > > > > > > the first job stored on it. > > > > > > > > On Wed, Sep 13, 2006 at 10:04:47PM +0200, Kern Sibbald wrote: > > > > > I haven't looked at the code, but I believe it is based on the > > > FirstWritten > > > > > which is the time (if I am not mistaken) when the Volume first had > data > > > > > written to it (i.e. just labeling it doesn't count). > > > > ... > > > > > No, I don't believe that is correct. The FirstWritten should be the > time > > > the > > > > > first non-label byte is written to the Volume. The whole process is > > > rather > > > > > complicated though. > > > > > > > > Nope. Here's the volume in question that started it all > > > > llist volumes output: > > > > > > > > MediaId: 4 > > > > VolumeName: clients-0004 > > > > Slot: 0 > > > > PoolId: 5 > > > > MediaType: File_clients > > > > FirstWritten: 2006-09-10 11:05:01 > > > > LastWritten: 2006-09-10 23:26:29 > > > > LabelDate: 2006-09-10 11:05:01 > > > > > > > > FirstWritten says 11:05, but the volume wasn't Recycled until 11 hours > > > > later, and first written another half hour later (I walked away and > forgot > > > > to mount the device...) > > > > > > > > 10-Sep 11:05 backup0-dir: Start Backup JobId > > > > 258,Job=hostname.client.com.2006-09-10_11.05.00 > > > > 10-Sep 11:05 backup0-dir: Max configured use duration exceeded. > > > > Marking Volume "clients-0003" as Used. > > > > 10-Sep 11:05 backup0-dir: Pruning oldest volume "clients-0004" > > > > 10-Sep 11:05 backup0-dir: Pruning oldest volume "clients-0004" > > > > 10-Sep 11:05 backup0-dir: Pruning oldest volume "clients-0004" > > > > 10-Sep 11:05 backup0-sd: Job hostname.client.com.2006-09-10_11.05.00 > > > > waiting. Cannot find any appendable volumes. > > > > Please use the "label" command to create a new Volume for: > > > > Storage: "Dev_clients" (/bacula/clients) > > > > Media type: File_clients > > > > Pool: clients_Pool > > > > 10-Sep 12:05 backup0-dir: Pruning oldest volume "clients-0004" > > > > 10-Sep 12:05 backup0-sd: Job hostname.client.com.2006-09-10_11.05.00 > > > > waiting. Cannot find any appendable volumes. > > > > Please use the "label" command to create a new Volume for: > > > > Storage: "Dev_clients" (/bacula/clients) > > > > Media type: File_clients > > > > Pool: clients_Pool > > > > 10-Sep 14:05 backup0-dir: Pruning oldest volume "clients-0004" > > > > 10-Sep 14:05 backup0-sd: Job hostname.client.com.2006-09-10_11.05.00 > > > > waiting. Cannot find any appendable volumes. > > > > Please use the "label" command to create a new Volume for: > > > > Storage: "Dev_clients" (/bacula/clients) > > > > Media type: File_clients > > > > Pool: clients_Pool > > > > 10-Sep 18:05 backup0-dir: Pruning oldest volume "clients-0004" > > > > 10-Sep 18:05 backup0-sd: Job hostname.client.com.2006-09-10_11.05.00 > > > > waiting. Cannot find any appendable volumes. > > > > Please use the "label" command to create a new Volume for: > > > > Storage: "Dev_clients" (/bacula/clients) > > > > Media type: File_clients > > > > Pool: clients_Pool > > > > 10-Sep 22:08 backup0-dir: Recycled volume "clients-0004" > > > > 10-Sep 22:08 backup0-sd: Recycled volume "clients-0004" on device > > > > "Dev_clients" (/bacula/clients), all previous data lost. > > > > 10-Sep 23:09 backup0-dir: Bacula 1.38.5 (18Jan06): 10-Sep-2006 23:09:58 > > > > JobId: 258 > > > > Job: hostname.client.com.2006-09-10_11.05.00 > > > > Backup Level: Incremental, since=2006-09-09 11:05:01 > > > > Client: "hostname.client.com-fd" i386-portbld- > > > > freebsd6.1,freebsd,6.1-STABLE > > > > FileSet: "hostname.client.com-fileset" 2006-07-09 > > > > 09:03:40 > > > > Pool: "clients_Pool" > > > > Storage: "Disk_clients" > > > > Scheduled time: 10-Sep-2006 11:05:00 > > > > Start time: 10-Sep-2006 11:05:01 > > > > End time: 10-Sep-2006 23:09:58 > > > > Priority: 10 > > > > FD Files Written: 699,861 > > > > SD Files Written: 699,861 > > > > ...etc > > > > > > > > Sorry, I don't have time to investigate the algorithm, but I did have > time > > > > to show the evidence as to what is happening here... > > > > > > I'm sorry, but this is not at all clear. I see what you have listed, but > > > without knowing what you have set for retention periods (if it is Volume > > > retention, what does the volume really have?, ...) and what kinds of > > > "forcing" of Bacula you may be using (Prune oldest volume, ...), I cannot > see > > > any evidence of a problem. > > > > > > Once Bacula asks for operator intervention, it is going to block until > > > you > > > specifically do a mount, unless you have polling turned on. It does not > set a > > > timer and then magically wake up at just the moment that the next > available > > > volume could be recycled. As I mentioned before this is very complicated. > > > > I have no problems with retention periods, recycling or mounting -- these > > are all working perfectly. Did you lose track of the topic? > > (not meaning to be rude, just confused) > > No, I have not lost track at all. The start time for Volume prunning is the > FirstWritten, that is clear, and you were complaining about Volume pruning > and recycling. > > > > > The issue is that the "FirstWritten" time on the volume is the start time > > of the job, not the actual time the volume was opened for writing. > > FirstWritten is not when the volume was opened for writing, it should be when > the volume was first written. > > > > > Job starts at 11:05 -- blocked waiting for a volume > > Volume recycled at 22:08 > > First job written to it at 23:09 > > > > Therefore, FirstWritten should be 23:09, not 11:05. > > That is not what you showed above (or at least it was not clear to me). I saw > a listing of the volume that I had to assume was before the job was run by > the wording of your sentence explaining the listing. > > If you are saying that FirstWritten is not updated when a volume is recycled, > that is possible, and if that is the case, it is probably a bug (this depends > on what FirstWritten means -- i.e. first time ever written or first time > written after a recycle), but I have never heard the problem phrased in that > manner, nor seen any examples (i.e. proof) of it. Is this the real problem? No, I'm not saying that. If that was true, the Firstwritten would be a month earlier.
I'm saying exactly what I've been saying in every post on this topic. * FirstWritten should be when first bytes are written to volume * FirstWritten currently _IS_ the Start Time of the first job written And please scroll up. You are looking at the proof. It's right there. -- Jo Rhett senior geek SVcolo : Silicon Valley Colocation ------------------------------------------------------------------------- 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
