Hi,
I've recently seen Amanda fail to write a dump to tape fully (write
error), but I wasn't prompted to run amflush to flush the remaining
files to tape, I usually am when there is data in the holding disk.
I'm not sure how this could happen, and this dump had several special
conditions. Anyways, I have configured amanda to flush the leftover
files at the next run. So she did, and hence there was way too much data
for a tape and it failed again.
>2.5 GB in the holding disk and the tape manages about 3.2 GB).
I am attaching the notification mails in order, host names have been
normalized and uninteresting gtar complaints ("socket ignored") omitted.
The interesting questions:
1. why did not amanda prompt me to run amflush on the first run, if
there was a tape write error?
2. why did not amanda massively bump to higher dump levels when there
was a file in the holding disk that had to be flushed first?
Because someone forgot to write the necessary code? :-)
Ultimately, a manual amflush run cleared everything up but I am under
the impression amanda should have asked me at the first dump to run
amflush. There has been a window of two days and the backed up file
system and the holding disk are from/on the same host. (I might not have
cared even half as much if the holding disk were in a different host.)
TIA,
--
Matthias Andree
Encrypt your mail: my GnuPG key ID is 0x052E7D95
l1
Description: first amdump, Feb. 14
l2
Description: second amdump, Feb. 16
l3
Description: amflush after amdump on Feb. 16
|