On Fri, Oct 17, 2003 at 11:33:45AM +0530, Rohit wrote:
> 
> I did few posts before expressing my concern about amanda not
> backing up few partitions. I realized that one of the reason
> could be that I put runspercycle as 5 weeks instead of 10 
> runs before. 
> 
> I guess amanda was thinking that it had approx 35 days to get 
> full dumps of each partition instead of 10 days. Is this right
> assumption? I made the correction yesterday - made runspercycle

I don't think your assumption is correct.  I think runspercycle
"ignores" the units part.  So it was thinking you had a total of
5 runs per dumpcycle.

> to 10 and removed 'weeks' from that line. I did this change on 
> the 8th day of my 10 day dumpcycle. I don't think amanda will
> be able to do full backup of those partitions in just 2 more 
> days left of dumpcycle. 
> 
> Or do I have to start fresh? (will amanda take care of this 
> policy change in between?)

You should not think of a dumpcycle as starting on a particular
date and going for one dumpcycle and the next not starting until
dumpcycle number of days.

Instead, each amdump run is the start of a dumpcycle's worth of
dumps, and can be thought of as the end of a different dumpcycle,
and can be thought of as an in-between run of other dumpcycles.
If you are doing 5 runs per cycle, tapes 1-5 contain "a dumpcycle".
But so do tapes 2-6, and 3-7 etc.

Using that kind of "mentality", amanda can easily adjust to policy
changes.  During its planning phase of an amdump, it simply
considers past history in the context of the new policies.

> Do you think I should wait for another cycle to complete and 
> raise alarm if those partitions still don't show up in the 
> backups? 
> What do you all say?

No, I think you have a problem there.  A problem that has little,
if anything, to do with the policies you were using.

I suspect things like:

  - those DLE's are simply too big to ever fit on a single tape,
    an absolute requirement for amanda to do the initial, and
    future, level 0's

  - a connection problem to those systems if all DLE's are failing
    for particular hosts

Time to get deeper into the reports, logs, and in /tmp/amanda, debug files.

> PS: Meanwhile I'm also thinking to force full backups of those 
> partitions which amanda has been ignoring. 

VERY unlikely it was ignoring them if they were in the disklist.

-- 
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)