Amanda twiddles its thumbs indefinitely when tape & holding disk are both full

Bug?

When the tape and the holding disk are both full, Amanda will just wait forever. I have to come in 
and kill it with `amcleanup -k daily`. My runs are typically running in the direction of 24 hours on 
one of the servers. So, I often leave at the end of the day assuming the Amanda run will complete 
shortly. However, sometimes a DLE will look like it's approaching 100% and in fact rather than 
finishing shortly, it runs on and on to 200% or 300%. Then everything stalls with no place to 
shuffle data. The next backup comes around and can't run, so I miss a day's backup. On a long, busy 
weekend, if I don't get a chance to check on it, I could lose a couple of days.

I'm running 3.3.9 on this server. Ubuntu 14.04, LTO6 and two 1T holding disks (SSDs actually). 
Compression is very good on a lot of these data sets, so Amanda may be pulling down 10TB or more of 
data onto that space on a daily basis with a 7 day dump cycle. Losing a couple of days can really 
throw a monkey wrench into the scheduling.

When I do an amstatus in this situation, I'll see several DLEs waiting for tape and a couple waiting 
for holding disk. The tape has filled so nothing can go to tape, and the holding disk has filled so 
nothing further can go to holding disk. In this situation, Amanda should do it's own termination and 
cleanup.


-- 
---------------

Chris Hoogendyk

-
    O__  ---- Systems Administrator
   c/ /'_ --- Biology & Geosciences Departments
  (*) \(*) -- 315 Morrill Science Center
~~~~~~~~~~ - University of Massachusetts, Amherst

<[email protected]>

---------------

Erdös 4