[Bacula-bugs] [bacula 0000908]: Tape not mounted at poll

[<bac...@li...>]

The following issue requires your FEEDBACK. 
====================================================================== 
http://bugs.bacula.org/view.php?id=908 
====================================================================== 
Reported By:                mschmid
Assigned To:                
====================================================================== 
Project:                    bacula
Issue ID:                   908
Category:                   Storage Daemon
Reproducibility:            always
Severity:                   trivial
Priority:                   normal
Status:                     feedback
====================================================================== 
Date Submitted:             07-28-2007 04:03 EDT
Last Modified:              07-28-2007 05:43 EDT
====================================================================== 
Summary:                    Tape not mounted at poll
Description: 
I have several DDS-3 tapes of different brands: HP, Certance and Compaq.
They may all be the same inside.

Bacula is up and running and it does its job well except for the tape
change.
To replace a veritas backup, I need a solution that is free of console
intervention. I've set up 'automatic mount', 'Offline on unmount', 'close
on poll', 'always open' and 'poll interval' so that a tape should be
reconized and mounted. 

The closest I get is that the tape is ejected when a new one is needed,
but that new one is not recognized. I see that the polling works halfways:
the drive goes active once long after tape insertions but bacula keeps
saying 'device open but no bacula media ....' 

- If I use 'always open = yes', it does not work. The tape is not mounted
on poll and I need to mount it using bconsole.
- If I have 'always open = no', it works fine and mounts the tape
automatically at its polling but it ejects the tape  at the end of each
job.

I've been thinking about a work-around using a cron script that would
mount the new tape thru bconsole if needed.

I've now modified the source so that the 'close on poll' really does an
unmount (mount.c: dev->unmount(0) instead of dev->close()), and it seems
that this is what I need to operate my drives. The patch is fresh and as a
test I've mounted the wrong tape manually and bacula ejected it, wanted the
right one and detected and mounted it properly.

At the moment I'd suggest an additional 'unmount on poll' feature

====================================================================== 

---------------------------------------------------------------------- 
 kern - 07-28-07 05:43  
---------------------------------------------------------------------- 
As best I can tell, by modifying the close() to be an unmount(), you
essentially turned it into a noop, which means (I think) you effectively
turned off the Offline on unmount command.

I suggest you try removing the Offline on unmount command and your patch,
and it will *probably* do what you want.

 

Issue History 
Date Modified   Username       Field                    Change               
====================================================================== 
07-28-07 04:03  mschmid        New Issue                                    
07-28-07 05:43  kern           Note Added: 0002624                          
07-28-07 05:43  kern           Status                   new => feedback     
======================================================================