Ben Jarvis <[EMAIL PROTECTED]> writes:
> I recently upgraded from 2.4.1p1 to 2.4.2p1, and ever since I've
>had a funny problem. Several of the disks repeatedly say "no estimate"
>when amdump runs ...
Ben and I worked out this problem offline (thanks, Ben!).
Whenever amcleanup is run (which Ben says he did a lot during the
upgrade :-), it looks for anything in the holding disk and marks it
"force-no-bump" so the same level will be done on the next run. It does
this because we don't know for certain the image is complete and it's
safer to ask for the same one again.
But there was a bug in planner when the level was 0 and this flag is set.
During dump analysis, it might decide to do an incremental since it
really wasn't time to do another level 0, but earlier code did not ask
for a level 1 estimate, which lead to the "no estimate" error.
Ben tested the following, and I reproduced the problem and tested it
as well. The fix is in the current source tree, but in case you're
having this problem, you can either apply this patch:
http://cvs.sourceforge.net/cgi-bin/cvsweb.cgi/amanda/server-src/planner.c.diff?cvsroot=amanda&only_with_tag=amanda-242-branch&r1=text&tr1=1.76.2.17&r2=text&tr2=1.76.2.18&r2=text&f=u
Or you can use amadmin to "unforce-no-bump" the affected clients/disks
(although you might want to also "force" them to make sure you have a
valid level 0).
It's still possible for planner to decide to do an incremental and thus
to go down the "no estimate" path, but only if the estimate of the level
0 also failed, in which case you're in deep doo-doo anyway :-).
>Ben
John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
