[OpenAFS-devel] Attempting to build a test environment. Help needed

[Marcus Watts]

> Date:    Mon, 19 Apr 2010 00:26:39 EDT
> To:      Marcus Watts <mdw@umich.edu>, Derrick Brashear <shadow@gmail.com>
> cc:      Mohammed Gamal <m.gamal005@gmail.com>, openafs-devel@openafs.org
> From:    Doug Hirsch <dhirsch@pobox.com>
> Subject: Re: [OpenAFS-devel] Attempting to build a test environment. Help neede
>      ***d
> 
> Derrick, Marcus,
> 
> Do you have a favorite public cheat sheet for setting up a test cell?
> 
> Doug
> 
> On Sun, Apr 18, 2010 at 11:17 PM, Derrick Brashear <shadow@gmail.com> wrote=
> :
> > On Sun, Apr 18, 2010 at 3:49 PM, Marcus Watts <mdw@umich.edu> wrote:
> >> I imagine anybody serious about afs development has a canned process
> >> for making a test cell, but I notice nobody posted theirs.
> >
> > I do it by hand every time.
> >
> > Your imagination is very vivid.
> >
> >> Bringing up a test cell is easy. =A0Do not let anybody tell you it's har=
> d
> >> or not worth doing. =A0You will learn so much more doing it that you can=
> not
> >> afford to miss the opportunity. =A0This is the difference between being
> >> a half-assed developer, and a kick-ass developer.
> >>
> >> You need 2 machines:
> >>
> >> =A0 =A0 =A0 =A0machine A
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0db server, fs server
> >> =A0 =A0 =A0 =A0machine B
> >> =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0test client
> >
> > You can do it with one if you're adventurous, crazy, or poor.
> >
> > If you do, make sure you configure the fileserver to start before you
> > configure the client to start.

Apparently Derrick has it hard-coded in meatware.
I assume he'll mail you a piece of his brain.  :-)

I don't install from scratch quite so often, so I rely on
rotating magnetic media to back my all too fallible memory.

I think I already mentioned
/afs/umich.edu/user/m/d/mdw/wp/uniq.2m
which should be publically readable.  It has the following limitations
and defects:
/1/ MIT kerberos only.  Heimdal should be conceptually the same.
	Recent MIT has bin/kadmin not sbin/kadmin.
/2/ pathnames specific to umich.  Any FHS host should be similar
	except for k5 paths which should be trivially convertible.
	Transarc based afs distro will differ more.
/3/ directions say to use pt_util.  This will work, but is
no longer best.  Today, in place of it, just below, after doing
	"bos create ... ptserver ...", do
	pts createu admin.admin 1 -localauth
	pts adduser admin.admin system:administrators -localauth
	(probably true for openafs 1.4.7 or better.)
/4/ don't build or run fakeka | krb524d
/5/ these notes say to run kadmind/krb5kdc out of bos.	This is not
	a requirement, but if you don't, you should have some init
	script that runs them.
/6/ make sure klog == klog.krb5 on client machines.
/7/ ignore any mention unserver, that's specific to umich.
/8/ in place of "take pre-built CD-rom contains virgin.tgz",
find an afs client, and using existing afs filespace where you
have permission to write, follow an adaption of the directions in
/afs/umich.edu/user/m/d/mdw/wp/hdserver.9
note: you don't need dogs.etc range.etc,
your fileserver is probably not named tomhorn, your cell is probably
not named range.  you should otherwise feel free to vary cell structure
to conform to local custom.  The filename ~/build/comma/,pwout
actually means
/afs/umich.edu/group/itd/build/mdw/comma/,pwout
but you only need that if you want home directory hashing.
Replace the pathname /afs/umich.edu/user/m/d/mdw/test with the path to
your writeable afs space.
/9/ The notes describe setting up and running the afs backup suite.
You can skip that if you don't intend to use it for backups.  You might
want to run "vos backupsys" periodically anyways.
/10/ these directions don't provide anything in DNS to help zero-configuration
clients.  You might want dynroot.
/11/ extra spelling mistakes.

As Russ points out, Debian provides very nice automatic cell creation
scripts.  They're great, especially if you don't want to know how the
pieces work, or you're already familiar so don't need a reminder.
Admittedly, knowing how to use kdb5_util is probably only going to contribute
to brain sludge formation.

				-Marcus Watts