More ipropd Anecdotes

["Henry B. Hotz" <hotz@jpl.nasa.gov>]

I have a cron job that does a stupid number of commands of the form

> kadmin -K <file> -p <admin> get <user>

Approximately twice a month I get messages in the admin log file on  
the slaves like:

> 2007-08-27T01:23:08 kadm5_log_replay: 108330: Entry already exists  
> in database

For a specifically bad incident there were 19511 kadmin get commands  
from 01:00:08 to 01:32:38.  The messages on the slave logs were from  
00:58:16 to 01:23:08.  No time sync problems I'm aware of.  The  
numbers of "exists" log entries were different on each slave (though  
the time span was the same), and different from the number of "get"  
operations as well.

12364 on kerberos02
14280 on kerberos03
  7204 on kerberos04

The status reporting on the master indicated that all slaves had  
returned to normal prior to 01:27:00.  kerberos02 iprop failed prior  
to, and was restarted at 02:15:57.

Obviously a "get" operation should not trigger a database update.   
However I can't push this point too far since there's only a general,  
not a specific correlation between the errors and the "get" operations.

I don't know if this information is of any use.  I know nothing  
specifically correlates.  Not sure if I can get more information, but  
feel free to ask.

Heimdal 0.7.2 with patches as discussed at the BP Workshop at SLAC.

------------------------------------------------------------------------
The opinions expressed in this message are mine,
not those of Caltech, JPL, NASA, or the US Government.
Henry.B.Hotz@jpl.nasa.gov, or hbhotz@oxy.edu