On Thu, 2006-02-09 at 22:10 +0100, Wolfgang Denk wrote:
> In message <[EMAIL PROTECTED]> you wrote:
> > 
> > In many cases (i.e. when, during the mv operation, to time stamp of the 
> > testfile inode was modified) /dir1/testfile will not be stored because 
> > it's not recognized as new - it's time stamps are older than the last 
> > full backup.
> 
> AAAARGHHH!!!! It was exactly reasons  like  this  one  that  made  me
> switch  from  my  home-grown  tar  scripts  to  a professional backup

Not to be curt, dotty, or any of that, but if If we could please move
the discussion back to "Testing and Simulation of the Scheduler", as
well as:

- Testing of Retention Times
- Testing of Recycling/Pruning
- Testing of automatic promotion of {Increm/Diferer}ential to Full...

How often does the Director poll the system date time?  How often does
the Storage director? I don't see any options in the config to control
this.  Specifically, for periodic maintenance like
recycling/pruning/volume status flagging.

Poll'ing for job schedule run time seems to be down to the second or
very close.

...let me give an example of behavior I'm seeing that seems
undesirable: 

I create a new file volume in a Pool "Daily" with a "Max Used Duration"
= 1h (3600 seconds), per my previous message "", *and* a Retention
threshold of 2 Hours.  This is for testing purposes only.  Here is the
llist output:

llist media pool=Daily        
          mediaid: 8
       volumename: CFusionDaily0
             slot: 0
           poolid: 2
        mediatype: File
     firstwritten: 
      lastwritten: 
        labeldate: 02/12/2006 18:40:56
          voljobs: 0
         volfiles: 0
        volblocks: 0
        volmounts: 0
         volbytes: 1
        volerrors: 0
        volwrites: 0
 volcapacitybytes: 0
        volstatus: Append
          recycle: 1
     volretention: 7,200
   voluseduration: 3,600
       maxvoljobs: 0
      maxvolfiles: 0
      maxvolbytes: 0
        inchanger: 1
          endfile: 0
         endblock: 0
         volparts: 0
        labeltype: 0
        storageid: 4


I start a single job manually using bconsole at:

Sun Feb 12 18:43:44 EST 2006

I get:

12-Feb 18:45 mindwipe-dir: No prior Full backup Job record found.
12-Feb 18:45 mindwipe-dir: No prior or suitable Full backup found. Doing
FULL backup.
12-Feb 18:45 mindwipe-dir: Start Backup JobId 62,
Job=Blackout-CFusionDefaults.2006-02-12_18.45.11

12-Feb 18:45 mindwipe-sd: Wrote label to prelabeled Volume
"CFusionDaily0" on device "FileStorage" (/usr/dump/bacula)
12-Feb 18:45 mindwipe-dir: Bacula 1.38.5 (18Jan06): 12-Feb-2006 18:45:22


List Jobs shows:

  Backup Level:           Full (upgraded from Incremental)
  Pool:                   "Daily"
  Storage:                "FileStorage"
  Scheduled time:         12-Feb-2006 18:45:00
  Start time:             12-Feb-2006 18:45:13
  End time:               12-Feb-2006 18:45:22
  Termination:            Backup OK



A success.  Afterwords, the llist status is:

*llist media pool=Daily
          mediaid: 8
       volumename: CFusionDaily0


     firstwritten: 02/12/2006 18:45:13
      lastwritten: 02/12/2006 18:45:22
        labeldate: 02/12/2006 18:45:13

        volstatus: Append
          recycle: 1
     volretention: 7,200
   voluseduration: 3,600


Now, theoretically, the Tape/Volume should remain status "Append" until:
02/12/2006 19:45:22

So I bump the date on the system:

# date 200602121947.00
Sun Feb 12 19:47:00 EST 2006

If I llist the volume:

     volstatus: Append

...remains.  If I "#pkill bacula"

It remains.  If I restart the Bacula daemons, it remains.

The only way I'm able to get the status to change to "Used" is if I do
in bconsole "status dir":


You have messages.
*message
12-Feb 19:50 mindwipe-dir: Max configured use duraVolume Poll Interval =
timetion exceeded. Marking Volume "CFusionDaily0" as Used.


However, the recycle flag remains, which I guess is okay (static flag,
doesn't change with recycle/prune/max* logic) because the logic won't
recycle the volume until Status changes back to "Used".

Now, If I bump the time ahead another hour to 20:50:

# pkill -HUP bacula

I immediately get:

*12-Feb 19:50 mindwipe-dir: Max configured use duration exceeded.
Marking Volume "CFusionDaily0" as Used.

Which is concerning, because I should see that tape getting marked as
pruned/recycled!

Once again, only until/after I run status dir:

====
You have messages.
*messages
12-Feb 20:51 mindwipe-dir: Pruned 1 Job on Volume "CFusionDaily0" from
catalog.
12-Feb 20:51 mindwipe-dir: Recycled volume "CFusionDaily0"

        mediatype: File
     firstwritten: 12/31/1969 19:00:00
      lastwritten: 02/12/2006 18:45:22
        labeldate: 02/12/2006 18:45:13
          voljobs: 0
         volfiles: 0
        volblocks: 0
        volmounts: 1
         volbytes: 1
        volerrors: 0
        volwrites: 231
 volcapacitybytes: 0
        volstatus: Recycle

# date
Sun Feb 12 20:52:36 EST 2006


So it seems that the director/SD only seems to process these time-based
directives after a 'status dir' is run.

~lava

Attachment: smime.p7s
Description: S/MIME cryptographic signature

Reply via email to