[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: snmpconf General Functional Questions - Policy GroupsandPriority

on 07/13/2000 2:35 AM, Steve Waldbusser at waldbusser@nextbeacon.com wrote:

> Jon Saperia wrote:
>> pmPolicyPrecedence OBJECT-TYPE
>> SYNTAX      Integer32 (0..65535)
>> MAX-ACCESS  read-create
>> STATUS      current
>> "The order in which policies on the local system are
>> evaluated. A policy with a higher precedence value will
>> be evaluated after a policy with a lower precedence. For
>> example, a policy with a precedence value of 999 will be
>> evaluated after a policy with a precedence value of 998.
> I don't think "time order" semantics will work very well. First of all,
> it seems to assume that there are no side effects of a variable going
> through a sequence of lower-precedence values before settling on the
> highest-precedence value. Second, we talked at the last meeting about
> having objects that control the frequency of policy evaluation on a
> per-policy basis, so it doesn't seem right to assume that all policies
> are evaluated at the same time, in order. Third, I'm concerned about
> what happens when an element changes state and suddenly matches
> a policyFilter - are we forced to rerun all higher precedence
> policyActions? Finally, consider that a policyAction may affect multiple
> variables - I think we want all of the variables to be set or none -
> it wouldn't be cool to change only those that survive higher precedence
> policies.

I did not reproduce the entire note here since the main points are above. In
short, we are in agreement that we want to have objects that control the
frequency of policy evaluation. I do not think what I propose requires that
all policies be evaluated in order. The precedence value will work even if
one policy is evaluated every minute while another that potentially
conflicts with it is evaluated only once a day. Once evaluated, the
information needed by the system to work is present.

We can talk more in Pitts.