Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/3/84; site maynard.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!allegra!mit-eddie!genrad!panda!talcott!wjh12!maynard!campbell
From: campbell@maynard.UUCP (Larry Campbell)
Newsgroups: net.unix-wizards
Subject: Re: Disappearing disk space
Message-ID: <147@maynard.UUCP>
Date: Wed, 4-Sep-85 23:23:11 EDT
Article-I.D.: maynard.147
Posted: Wed Sep  4 23:23:11 1985
Date-Received: Fri, 6-Sep-85 03:17:50 EDT
References: <1210@brl-tgr.ARPA>
Organization: The Boston Software Works Inc., Maynard, MA
Lines: 34

I've been getting the same few suggestions repeatedly in response to
my original question, so here's what I should have posted earlier --
a summary of hypotheses people have offered that have not panned out:

(Synopsis: VENIX/86 system mysteriously loses many thousands of blocks
	   of disk space which later mysteriously come back.  Latest
	   discovery is that deleting /usr/lib/news/history frees up the
	   space, even though that file claims to be fairly small.
	   VENIX/86 is a V7 port;  news software here is 2.10.2.)

Several people have suggested a temporary file that gets unlinked
but not closed.  Nope, reloading doesn't free up the space.

Another suggestion was a filesystem mounted on a nonempty directory.
Very plausible, but not the case here.

A third suggestion was holes in a file (lseek past EOF).  I wrote a
test program to try this out, and only succeeded in getting du and
quot to think that the file occupied MORE blocks than it really did.
My problem is the reverse -- that du and quot think the filesystem has
lots of space, but it really doesn't.

Since du seems to believe the EOF pointer (st_size field of stat(2)),
it's occurred to me that perhaps /usr/lib/news/history is somehow
getting blocks allocated past EOF.  I can't think of a way to make
that happen, though.  Any suggestions?
-- 
Larry Campbell                     decvax!genrad
The Boston Software Works, Inc.                 \
120 Fulton St.                 seismo!harvard!wjh12!maynard!campbell
Boston MA 02109                         /       /
                                   ihnp4  cbosgd

ARPA: campbell%maynard.uucp@harvard.arpa