check disc space

[David M. Wilson]

"Fredrik Lundh" <fredrik at pythonware.com> wrote...

> - the os module is traditionally used to provide access to posix-style
> services provided by the OS (or the C library).  in this case, os.statvfs
> is such a function; calling out to "du" isn't

> (shouldn't that be "df", btw? os.popen("du -k dir").split()[-5] sure
> doesn't give the free diskspace on any platform I have access to...)

Um, yep. Fixed. *sheepish shrug*


> (a third argument is that "free disc space" is a rather meaningless metric
> on a modern operating system; it's hardly ever useful for anything except
> "warning, warning, less than 5% free space" alerts.

Considering what I see python used for on a day-to-day basis, this
isn't "hardly ever useful". I have seen jobs fail as a direct result
of no disk space check. This should indeed be handled elsewhere and in
a different fashion, but sometimes the world just ain't the ideal
place you'd like to believe it is.

When you are performing 4gb deduplication runs on a cluster, and have
to wait a couple of days only to find out you don't have enough disk
space to write out the last 5 records of the result tables, I think
this silly little module can be justified.

And why not just use the POSIXy interface provided by os? Well, the
company I work for uses the same codebase for both Windows and Linux
clustering, which is a quite a feat. All support tools as such must
run under both operating systems too.


> you really don't want
> people to write "if os.getfreespace() >= len(buffer): f.write(buffer)" and
> file a bug when write fails ;-)

Certainly not! :)

But "if estimated_space > int(free_space * 1.05): ..." might not be
such a bad thing when there is no other route to sanity available.

Thanks,


David.