Any idea

[Marcus Watts]

Terry McCoy <terry@nd.edu> writes:
> Date: Thu, 24 May 2001 20:31:37 -0500 (EST)
> From: Terry McCoy <terry@nd.edu>
> To: info-afs@transarc.com
> Subject: Any idea
> 
> Below are messages that were recorded in the FileLog on a number of
> my AFS servers (one message per server).
> 
> 
> Would anyone be able to tell me what these messages mean?
> 
> 
> Thu May 24 13:18:16 2001 Host 65fee0 used to support WhoAreYou, deleting.
> Thu May 24 13:18:18 2001 Host 419dc0 used to support WhoAreYou, deleting.
> Thu May 24 13:21:44 2001 Host 837af0 used to support WhoAreYou, deleting.
> Thu May 24 13:21:47 2001 Host 849508 used to support WhoAreYou, deleting.
> Thu May 24 13:21:48 2001 Host 41cce0 used to support WhoAreYou, deleting.
> Thu May 24 13:21:48 2001 Host 80bb58 used to support WhoAreYou, deleting.
> Thu May 24 13:21:48 2001 Host 9355e0 used to support WhoAreYou, deleting.
> Thu May 24 13:21:48 2001 Host 9dc880 used to support WhoAreYou, deleting.
> Thu May 24 13:44:24 2001 Host d762f8 used to support WhoAreYou, deleting.
> 
> 
> tHanKs
> --
> Terry McCoy                             email:  terry@nd.edu
> Sr Systems Engineer                     phone:  (219) 631-4274
> Office of Information Technologies
> University of Notre Dame
> 
> 

These come from viced/host.c, in the routine h_GetHost_r.  In AFS 3.4,
this routine was called h_GetHost, and it always made a call to
RXAFSCB_InitCallBackState when a new connection from a cache manager
was seen.  In AFS 3.6, there's apparently a replacement call,
RXAFSCB_WhoAreYou, and on the initial contact, this call is first
made.  If it returns RXGEN_OPCODE (ie, it was an "old" cache manager),
then it reverts to doing RXAFSCB_InitCallBackState and marks
identP->valid 0, otherwise, it marks identP->valid 1.  In otherwords,
it starts out assuming 3.6, and falls back to 3.4 behavior.  On
subsequent calls, if h_GetHost_r gets a RXGEN_OPCODE error, it
complains
	Host %x used to support WhoAreYou, deleting.
and arranges to start over.

The only reason you *should* be seeing such messages is if you had one
or more machines that was for some reason switching cache manager
versions fairly frequently, or if you had a really bizarre problem with
a router (or something doing some really wild dynamic network IP
address translation (NAT) with UDP datagrams)?

There's more--those numbers 65fee0 419dc0 837af0 849508 etc. don't look
like valid IP addresses.  In fact, it's the internal memory address
of a structure that points to host information, which means it's
practically useless for figuring out anything really useful about
the problem (like is it a host or router problem.)  Probably that line
ought to read:
	ViceLog(0,("Host %x.%d used to support WhoAreYou, deleting.\n",
                        host->host, host->port));
but there's a bunch of similar lines that need about the same fix,
and even more that log host->host, but don't bother to include host->port.
It would be even nicer if they all used ntohl on the host address,
or converted it to dotted decimal notation.

				-Marcus Watts
				UM ITCS Umich Systems Group