Bacula-users

Re: [Bacula-users] Incremental upgraded to full?

2008-04-26 13:21:21
Subject: Re: [Bacula-users] Incremental upgraded to full?
From: "Timo Neuvonen" <timo-news AT tee-en DOT net>
To: bacula-users AT lists.sourceforge DOT net
Date: Sat, 26 Apr 2008 19:49:29 +0300
"Alan Brown" <ajb2 AT mssl.ucl.ac DOT uk> kirjoitti viestissä
news:Pine.LNX.4.64.0804251101380.5996 AT mssllu.mssl.ucl.ac DOT uk...
> On Thu, 24 Apr 2008, Dan Langille wrote:
>
>> > Last night I ran the backup cycle, bacula did the full backup and
>> > immediately tried to do the incremental backup. Instead of doing
>> > incremental backup it upgraded it to a full backup. Why did it upgrade
>> > to a full backup if full backup was done right before it started
>> > incremental?
>>
>> The upgrade to FULL is done at the time the job is put into the queue.
>> At that time there was no FULL backup.
>
> IMHO this is a bug.
>
> The status check should be done when the job exits the queue and starts
> up, not when it's placed on the queue.
>

This could result in a major difference in behaviour depending on whether
the jobs are run concurrently or one-by-one, or if the number of jobs were
limited by maximum concurrency. Does not sound very nice.

I can't know what the administrator exactly wanted to do in this case, but
as fas as I can think I would consider the bug to be in the way how the jobs
were scheduled to run.

So, the next time there will be a one-month old full backup, and another
full backup being done. And at the same time with the full backup, a pending
incremental job that obviously is based on the old full backup. Here I would
see it irrelevant if it is based on the old full job, or the new one, though
the first choice would result in a larger job. Just why to run at all that
incremental job that most likely is expected to contain nothing, but the
contents of which significantly depend on the timing and implementation
details?

Why not to change the schedule so the incremental job would not be run at
all that night? Either by originally excluding it from the schedule, or
cancelling it with max wait time etc directives.


Regards,
Timo



-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
Bacula-users mailing list
Bacula-users AT lists.sourceforge DOT net
https://lists.sourceforge.net/lists/listinfo/bacula-users