Re: [Bacula-devel] Re: Compile errors

[Kern S. <ke...@si...>]

Well, I don't imagine you are still there. The problem
is that you have created something that I never imagined
would happen -- now that I see it, it is logical, but
Bacula sure doesn't know how to handle it. It is your
one giant file that spans more than two volumes.

In that particular case, there is nothing 100% reliable
that Bacula has to know the order of the JobMedia records.
I had always assumed that the start/end indexes (sequential
file number) would be sufficient, but here is a dump of 
the JobMedia records for your Job.

select * from JobMedia where JobId=61;
+------------+-------+---------+------------+-----------+
| JobMediaId | JobId | MediaId | FirstIndex | LastIndex |
+------------+-------+---------+------------+-----------+
|         37 |    61 |      33 |      74695 |     74696 |
|         36 |    61 |      32 |      74695 |     74695 |
|         35 |    61 |      31 |      74695 |     74695 |
|         34 |    61 |      30 |      74695 |     74695 |
|         33 |    61 |      29 |          1 |     74695 |
+------------+-------+---------+------------+-----------+

where I have chopped of a few columns to make it fit in
the email.

Notice that FirstIndex and LastIndex are identical for
JobMediaIds 34,35, and 36!  FirstIndex was the sort key.

I could sort on JobMediaId, which I will do as a temporary
stop gap, but IMO, it is a very *BAD* idea to count on any
database to *always* sequentially increment the primary key. 
Just imagine what happens after 4 billion entries when the 
Id wraps around.  

I need to add a new field to the JobMedia table which is
the sequential volume number within the job. Then Bacula
can be 100% sure of the order the JobMedia records were
created.

Best regards,

Kern



On Fri, 2003-05-23 at 22:55, Dan Langille wrote:
> On 23 May 2003 at 22:39, Kern Sibbald wrote:
> 
> > No, your Volumes are not correctly sorted.
> > Back to the drawing board ...
> 
> OK, give me the SQL or point to in within query.sql and I'll play 
> with it here.  I have about 20 minutes left.