Number of successful postings ....

John Shott shott at snf.stanford.edu
Tue Jan 16 15:13:49 PST 2001


Bill and Mike:

In looking at evtmgr.log, I'm seeing some strange behavior ... in particular,
if we believe that a posting that results in errors from either a maintenance
client, a reservation client, or an equipment client should result in a
cleanup of each subscription list so that all events should successfully post
to the same number of clients (except for things like problems and shutdown
that should post to 2x number of clients).  So, what I have been seeing is a
little wierd ...

Following are some annotated logs when I exercised bottlewash while the lab
was fairly heavily loaded:

Enable:
2001-01-16 14:44:19 Posted event: 101 (1461, 86, 0) Elapsed time: 9 + 701 msec
Disable:
2001-01-16 14:44:37 Posted event: 102 (1461, 86, 0) Elapsed time: 2 + 321 msec
Problem:
2001-01-16 14:45:13 Posted event: 104 (1461, 264, 1) Elapsed time: 3 + 2266
msec
Problem clear:
2001-01-16 14:45:53 Posted event: 106 (1473, 179, 0) Elapsed time: 40 + 378
msec
Problem clear all:
2001-01-16 14:45:53 Posted event: 108 (1473, 86, 0) Elapsed time: 471 + 333
msec
Reservation:
2001-01-16 14:46:46 Posted event: 201 (1473, 118, 0) Elapsed time: 8 + 773
msec
Reservation delete:
2001-01-16 14:46:52 Posted event: 202 (1473, 118, 0) Elapsed time: 9 + 770
msec
Shutdown:
2001-01-16 14:51:03 Posted event: 103 (1455, 263, 0) Elapsed time: 2 + 576
msec
Shutdown clear:
2001-01-16 14:51:29 Posted event: 107 (1455, 178, 0) Elapsed time: 66 + 494
msec
Shutdown clear all:
2001-01-16 14:51:29 Posted event: 109 (1455, 84, 0) Elapsed time: 564 + 183
msec

If we believe that all clients are fully in sync, then I would have thought
that all events should have been reported to approximately 86 clients ...
well, except for problem, problem clear, shutdown, and shutdown clear that
actually get reported to both the maintenance clients and the equipment
clients.

The strangeness that I see is that problems and shutdowns seem to go to about
3x the number of active clients whereas I would have thought they should go to
2x the number of active clients (mainenance client and the equipment client). 
It makes me think that we are not handling subscription to the problem and
shutdown events properly and that either the maintenance client or the
equipment client is double subscribing.

Also, the number of reservations postings is rather wierd ... 118 instead of
86 or so.  I don't know what this means ...



More information about the coral mailing list