On Fri, Sep 02, 2005 at 08:50:21PM -0400, Joshua Baker-LePain wrote:
> On Fri, 2 Sep 2005 at 3:00pm, Joe Rhett wrote
> 
> > On Thu, Sep 01, 2005 at 05:06:37PM -0400, Joshua Baker-LePain wrote:
> > > *You* are the only one who can really identify the problem.  Are you 
> > > using 
> > > server-side or client-side estimates?  If the latter, look at the 
> > > sendsize*debug for the 53GB dump to see what amanda thought the size 
> > > would 
> > > be at estimate time.
> >  
> > See the attached log except.  Amanda knew it was bigger than a tape, but
> > backed it up anyway.
> 
> I'd be intereseted in seeing the sendsize*debug from the client and/or 
> the relevant planner line from the amdump.N file.  Amanda didn't do a 
> level 0 b/c it knew that was bigger than tapelength, but then the level 1 
> turned out to be even bigger than the estimate for the level 0.  ??
 
Here's the relevant lines from amdump's log:

planner: time 0.084: setting up estimates for customer-plat1:/
customer-plat1:/ overdue 82 days for level 0
setup_estimate: customer-plat1:/: command 0, options:
    last_level 1 next_level0 -82 level_days 9
    getting estimates 0 (10) 1 (-1) -1 (-1)

planner: time 864.871: got result for host customer-plat1 disk /: 0 -> 
53227860K, 1 -> 53195890K, -1 -> -1K

pondering customer-plat1:/... next_level0 -82 last_level 1 (due for level 0) 
(picking inclevel for degraded mode)   pick: size 53195890 level 1 days 9 
(thresh 204800K, 1 days)

INITIAL SCHEDULE (size 145819677):
  customer-plat1 / pri 83 lev 0 size 53227860
        ...etc

DELAYING DUMPS IF NEEDED, total_size 145819677, tape length 241282048 mark 32
  delay: customer-plat1 / 20050825 0 [dump larger than tape, 53227860 KB, full 
dump delayed] now at level 1
  delay: Total size now 145787707.

DUMP customer-plat1 34cbfe811f0100 / 20050825 83 1 2005:5:28:8:23:38 53195890 
7092785

The only remaining lines referecing it are the flush, then taper, then
taper-tryagain... repeat twice more.

-- 
Joe Rhett
senior geek
meer.net