Coda File System

Re: newbie questions re: coda media server

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Mon, 26 Jan 2004 10:38:56 -0500
On Sat, Jan 24, 2004 at 09:38:26AM -0500, Adam Manock wrote:
> I have an MS Xbox running Xebian 1.0 (Debian for Xbox)
> I have successfully installed coda client 6.03-1
> (It doesn't get easier than "apt-get install coda-client") 

That's really nice.

> I am now wondering what the preferred way to disconnect from the
> testserver is. It's not immediately apparent how to remove
> /coda/testserver.coda.cs.cmu.edu/ from the filesystem now that I'm done 
> playing with it. Does it go away on it's own?

It should go away on it's own, were it not for a bad refcount which is
already fixed in the CVS tree.

To normally speed up discarding the realm, you could use 'cfs fl
/coda/testserver.coda.cs.cmu.edu'. With the bad refcounting, this
doesn't work right in 6.0.3 either.

> I'm also wondering what the best options would be for running a coda
> media server. It would be serving mostly big files. 100Mb quicktime
> movies, 3-6Mb MP3s etc. 

One gotcha is the whole file caching, so for read-only operation you
need to have at least as much as your largest file as a Venus cache. If
you also want to write, probably double the size. And then add a bunch
for 'overhead', such as the directories in the path leading to the file
etc. So I would say, about 300Mb cache. But on the other hand as the
average file is pretty big, you really don't need all that many objects
cached locally, so you can keep the size of venus pretty mean and lean
by separately configuring the number of cachable objects.

i.e.
venus.conf
    cacheblocks=300000
    cachefiles=2500		# with a 300MB cache this is normally 12500

> Do I now need to create a coda filesystem (volume?) on my server and
> insert all my media into it? If I use my server as a workstation, do I
> have to use coda client/server protocols to access the same content 
> locally?

Yes, you would use a Coda client on your server. We don't export an
existing tree because Coda needs to be able to track changes to files.
That way it can handle replication and send callbacks to invalidate
cached copies when files are changed on the server.

Jan
Received on 2004-01-26 10:51:31