Jay wrote:
______________________________ Reply Separator _________________________________
Subject: Re: Re[2]: PMP> Should Alerts be replicated???
Author: jkm at underscore.com (JK Martin) at Internet
Date: 1/14/97 12:15 PM
Not reentering the warning doesn't feel right to me. Maybe it's just
a gut feeling, but without critical/warning alerts in the Alert Table,
the mgmt app, upon initial contact with the printer, is forced to query
the subunit status of every component in the Printer MIB to determine
if any problem states exist.
Perhaps I should be clearer about why I don't suggest formalizing a
re-entry. (I do not suggest forbidding it either.)
Under what circumstances would a significant warning be removed from
the Alert Table?
By the suggested Alert Table maintenance, only if a more critical (or
more recent event of equal criticality) occurs and the table is
already full. Note that the maintenance suggests that non-critical
unary alerts can be supplanted by more recent non-critical unary
events or by noncritical binary events or by critical binary events.
It might also be suggested that the printer manufacturer can assign
some finer priority level among unary noncritical events, and among
non-critical binary events, etc. Therefore, an event that gets
supplanted will be the least important of all events in the Alert
Table. In addition, unless it is a less important byproduct of an
event causing many alerts, this supplanted alert event probably will
have resided in the Alert Table for some period and may well have
already been observed by the management application.
When would room be created in the Alert Table to allow the supplanted
event to be reentered?
Since all events in the Alert Table are of higher priority, slots will
be freed in the Table only if some binary alert has a terminating
event. Except for a critical binary event, which must be terminated
via operator intervention if the printer is to resume operation, if
and when the slot will be freed is not clear and not something a
Management app could rely upon. This unlikeliness of freeing up space
is greater if the alert alert option is used.
Therefore, I suggest that an event will be supplanted only if it is
the least significant event in the table, that the event probably will
have been in the table for some time, and that space to accept a
re-entry of that event would become available only if more significant
conditions were removed, probably by operator intervention.
Bill Wagner