>... I noticed this after discovering that the file created on the
>holding disk for that dump was "sabre._biff.7". ...
>...
>So why did amstatus report it as level 0? ...
Because it's a bug :-), or else it got bad information from the
amdump.<NN> file.
As often as I've mentioned it, I'm not sure everyone understands amstatus
is a highly inexact representation of things. The main issue is that
almost all of its data comes from the amdump.<NN> file, and several
processes are writing to that with no locking. Messages get lost and
corrupted in there all the time, which, of course, confuses amstatus to
no end.
>If I had let the amdump
>finish, I would be curious to know what level would have actually been
>performed and what the final report would have stated.
You already noted that the holding disk file ended in .7, so Amanda
was doing a level 7. And that's what the E-mail (and database, etc)
would have recorded since that code is careful to do the right thing
(i.e. proper locks are used).
Note that I'm not against amstatus in any way. I use it all the time.
But I take what it tells me with a grain of salt (although I agree
reporting the wrong level number is worse than usual :-).
If you'll send me that amdump.<NN> file (I think that's all I need),
I'll see if the fix is within reason. It might just be that amstatus
doesn't understand degraded mode. More likely, though, is that some
key message it needed was corrupted and there isn't much it can do.
I have a TODO item to see if at least some minimal locking can be applied
to the amdump.<NN> file, which would help amstatus (and other debugging).
>Notice the "not due for a full dump, picking an incr level"
>paranthetical statement. Also note the "promote: moving
>sabre:/biff up, total_lev0"
That's just planner saying "it's not time to do a level zero so I'll pick
an incremental", then later saying "oops, I'm going to be in trouble
in the future if I let the level 0 happen when it's next scheduled,
so I'll try to move it forward to this run".
>Paul
John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
