Hi!
John R Jackson writes:
>> It's a Solaris 2.5.1 box. I didn't specify any chunksize, so my
>> thinking was that the default of 1 Gb is fine.
John> It should have been. You might set it explicitly just to be
John> sure.
I'll try that.
John> You might also check the ulimit values (and maybe the disk
John> quota) for your Amanda user to make sure you're not banging into
John> that kind of problem.
Ok, I'll check that, although my amanda user is root (probably not the
best idea) and I didn't explicitly set any limits.
>> However, thanks for the pointer! I'll play with a smaller chunksize,
>> especially a chunksize smaller than the `use - xxx Mb' value...
John> The "chunksize" and "use" values are not related, and
John> "chunksize" certainly does not have to be smaller than "use".
John> The "use" value determines how much total space Amanda has to
John> work with in the holding disk. The "chunksize" value tells it
John> how to split apart a single image.
That's how I did understand the manual, as well. I was just thinking
that there might be a problem with concurrent dumpers writing to
the holding disk and the negative notation of the `use' parameter.
Probably my next attempt will be to attach strace to a running dumper
for a filesystem that I `expect' to fail and concurrently observing
the holdingdisk space.
-frank
