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

Re: snmpconf General Functional Questions



On Fri, 9 Jun 2000, Jon Saperia wrote:

> > So maybe what we really want is a "Don't touch" table of locally
> > configured OIDs and the ability to delete OIDs from that table when the
> > local configuration is no longer relevant?
> 
> I think the 'don't touch' is an attribute of an element (instance) if it has
> been placed under policy control and has then been modified by something
> other than the policy system. The don't touch can only be helpful I think if
> it is that specific and the don't touch is associated with a policy. What do
> you think?

The problem is what happens if an element is locally modified and then put
under policy control?  I don't think we want to allow policies to
overwrite existing explicit (manual) configuration when they are first
instantiated.

The way I view this is that, at any given time, there will be some number
of policies being enforced on a device.  These policies will conflict with
eachother (that seems unavoidable).  We need a mechanism for resolving
this conflict regardless.  Now, if we view manual configuration as a
"policy" of the highest priority, the conflict resolution mechanism that
we already need can be used to give us the behaviour that we want.

So, what I suggest is a table of OIDs and RowPointers to the policies that
are modifying them (there can be multiple policies per OID).  Manual
configuration can be a special policy that always exists.  Each policy
then has a priority assigned to it that is used to determine which policy
of the conflicting policies take effect.


-Matt

Matt White
Ericsson IP Infrastructure