[2751] in Kerberos
Kerberos server incident report - not sure if it's a bug
daemon@ATHENA.MIT.EDU (Jim Haynes)
Tue Jul 20 14:43:55 1993
From: haynes@cats.ucsc.edu (Jim Haynes)
Date: Tue, 20 Jul 93 11:23:18 -0700
To: kerberos@Athena.MIT.EDU
The Kerberos server is running on a Sun 3/50 under SunOS 4.1.
There are about 7000 entries in the data base.
I tried kadmin and then get *.instance
where instance is an instance that happens not to be in the data base
at all. The server got awfully busy, being constantly in run or disk
wait state. After a while I got tired of waiting for the response
to kadmin and gave it a ctrl-C. The server kept on being busy; and not
knowing if it would ever stop I killed it and restarted. A stack trace
with adb shows it is in
fstat(?)
_DYNAMIC(
kerb_db_get_principal() called from
kerb_get_principal("*", "instance", something, 1, pointer to zero)
I'm not sure if the absence of arguments to kerb_db_get_principal() is
pathological; I'm no ace at adb.
And maybe the server would eventually have come out of its intensely
busy state, but while it was in that state it wasn't answering any
queries so we were dependent on the backup server.
Guess I'd like to know if anybody else has seen this behavior.
