Subject: Re: nullfs still doesn't reclaim space (PR#15555)
To: Jeremy C. Reed <reed@reedmedia.net>
From: Simon Burge <simonb@wasabisystems.com>
List: tech-kern
Date: 02/15/2003 12:03:57
"Jeremy C. Reed" wrote:

> In my last class, I accidently untarred the apache tarball on the wrong
> mfs, and when I deleted the files (and they weren't in use), my file
> system was still 101% full. So I rebooted, which caused another unrelated
> issue -- the overhead video projection wouldn't sync anymore :(

I forgot to mention that first time around that flushing the vnode cache
will eventually get your space reclaimed. "du -s <somewhere>" where
"<somewhere>" has more files that your kern.maxvnodes setting is the way
I do it.

In the "really crude but somewhat effective" category, however, I've
just discovered that there is a way of instantly clearing the vnode
cache.  The code to handle increasing the cache does:

			if (old_vnodes > desiredvnodes) {
				desiredvnodes = old_vnodes;
				return (EINVAL);
			}
			vfs_reinit();
			nchreinit();

so

	sysctl -qw kern.maxvnodes=`sysctl -n kern.maxvnodes

will flush the cache for you.  You can verify this by doing
"vmstat -h nchash" before and after the sysctl command.

Simon.
--
Simon Burge                            <simonb@wasabisystems.com>
NetBSD Support and Service:         http://www.wasabisystems.com/