From mboxrd@z Thu Jan  1 00:00:00 1970
Message-ID: <67f19d498e0308564756569409b13b07@plan9.bell-labs.com>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] df for fossil?
From: "Russ Cox" <rsc@plan9.bell-labs.com>
In-Reply-To: <7f141a65922d5fa74df962c7721f4a69@plan9.escet.urjc.es>
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Date: Tue, 14 Jan 2003 10:06:56 -0500
Topicbox-Message-UUID: 40f38ad6-eacb-11e9-9e20-41e7f4b1d025

When a partition gets full you get lots of disk full
errors, as you might expect.  Unlike in a normal file system, you
might not even be able to walk around and read the file system,
since walking the file system potentially means allocating
blocks to update atimes.

It recovers gracefully though.  When sources filled
a few weeks ago, I just connected to the console and
tossed some old snapshots with epoch -y.

If you fill the write buffer and have no snapshots to
discard, then all your blocks must be mutable, so 
you should be able to walk around in the file system
and remove files.  We should probably keep a few
blocks reserved for running an emergency flush-to-Venti 
`snap -a'.  (There's little hope for recording the snapshot
in the /archive tree, but if we kept one block around then
we could use it to clone the root and start an archive of
the resulting anonymous snapshot.)

Russ