Experimenting with milko

[Karl Kleinpaste]

Love <lha at stacken.kth.se> writes:
> Point 2 decribes how the principal afs at REALM gets created.

> If you failed to create the afs principal in step 2, you'll make the
> afs at REALM in /usr/arla/etc/srvtab get out of sync with the afs at REALM i the
> kerberos server if you add a new one.

Hm.  Well, OK, I went back and re-read it, and re-read it again, and
re-read it once more.  I'm very new to kerb and afs admin things, and
it really didn't occur to me that I had to create afs at REALM with
kadmin first -- I thought some implicit creation was going on when
doing ksrvutil (though I was concerned that I had no input as to the
"afs" principal's password).

Also, another source of 40 minutes of wasted time is not quite having
noticed the little "''" in the ksrvutil dialogue for "Instance" under
#2, for generating /usr/arla/etc/srvtab.  Because I missed that, I
inadvertently generated a srvtab for afs.host at REALM rather than
afs at REALM, again leading to "rxkad - Unknown key" errors.

Generally, it seems these paragraphs are too short, and could use some
expansion for the sake of those of us who are new to the problems at
hand.  I don't think I'm qualified to be the one writing the updates.

So, having gotten the issues of the "afs" principal handled well
enough (vos finaly worked), I tried running ptserver.  There are no
special instructions in milko/README; just run it.  This doesn't work
immediately -- fortunately, "ptserver -help" reveals another "-create"
option, and having done this first, a subsequent simple run of
ptserver seems to be doing something normal.

And then, having started fileserver as well, I seemed to have the
whole situation initialized and running properly.

Unfortunately, starting up arla itself, I now find that I can't cd
anywhere in /afs at all, and the fastest way I've ever seen to
generate hundreds or possibly thousands of XFS panic errors is to "cd
/afs" followd by just "ls".  This gives me boatloads of...

      Message from syslogd at mesquite at Tue May 23 22:13:02 2000 ...
      mesquite kernel: XFS Panic: xfs_message_installdata failed open_namei, errno: -2 
      Message from syslogd at mesquite at Tue May 23 22:13:02 2000 ...
      mesquite kernel: XFS Panic: xfs_message_installdata failed fh_to_dentry = 00/01@, errno: -2 

Fortunately, simply killing the ls(1) stops these.

Platform is Linux RH6.2, kernel 2.2.14.  If anyone has any wisdom to
share on why everything under /afs is simply AWOL after having started
the assorted servers, and why trying to cd to a known-working cell
(cs.cmu.edu) causes such trouble, I'd welcome suggestions.

--karl