On Tue, 2005-10-11 at 09:51, Paul S. Gumerman wrote:
Is the RsyncFileIO perl module using bad arithmetic for device space
calcs that can't handle multi-TB devices? Or does it just make the
assumption that the problem is lack of space?
The assumption is that if the OS returns an error for an
operation that the error number is accurate... Your error
wasn't about calculations, the mkdir attempt failed.
You didn't answer the request to run 'df -i' on the filesystem
in question. I think 'out of inodes' may return the same
error as 'out of space'.
I do have one observation/criticism after spending a lot of time going
through the list archives: it seems that the transfer processes for
backuppc are overly fragile --- if there is a problem, the whole
backup process stops in its tracks, rather than noting the problem and
trying like hell to get the rest of the files. Don't know if anything
more could be done to increase the robustness or not, and I'm hardly
an expert on backup systems --- so take this with a grain of salt.
The one problem you see a lot in the archives is that large
runs used to time out regularly because with some methods
the timeout was for the whole run, not for any activity.
This should be fixed with the much longer default in the
current release. If the OS refuses to create or write
files, there isn't much an application can do to fix it,
though.
