Rules For Attachment Through Interface Configuration Mode - Juniper E320 Configuration Manual

An interface and an attachment type identify an attachment point. The policies
referenced by the component attachments merge into a new policy, which then
attaches at the attachment point. The set of component policies are ordered
alphabetically by name. This order determines how any merge conflicts are
resolved, with the most recently executed command taking precedence.
With policy merging, a set of policies is combined to form a single new policy,
which is a union of all the component policies. Classifier groups and policy rules
from each component combine to create the merged policy as in the following
example:
host1(config)#interface atm 5/0.1
host1(config-subif)#ip policy input p1 statistics enable merge
host1(config-subif)#ip policy input p2 statistics enable merge
host1(config-subif)#ip policy input p3 statistics enable merge
host1(config-subif)#ip policy output p4 statistics enable merge
host1(config-subif)#ip policy output p5 statistics enable merge
host1(config-subif)#exit
The example internally results in the following, where policies p1 + p2 + p3 =
mpl_10 and policies p4 + p5 = mpl_11.
interface atm 5/0.1
ip policy input mpl_10 statistics enable merge
ip policy output mpl_11 statistics enable merge
exit
The classifier list referenced by the classifier group is neither split or merged. If a
merged policy already exists for a set of component policies, then the merged
policy is used for the attachment. An attachment enables a merged policy to have
one or more attachments.

Rules for Attachment Through Interface Configuration Mode

The CLI and the Service Manager applications are the only clients of policy
management that can request merging of policy attachments. With policy merging,
classifier groups and policy rules from each component policy combine into the
merged policy.
Policy merging follows these rules:
!
!
!
The Classifier list referenced by the classifier group cannot be split or merged.
Policy merging combines classifier groups from all component policies into the
merged policy. In the previous example, policies p1, p2, and p3 are the
component policies and mpl_10 is the merged policy. The merge policy is
created as if all CLI commands for each component policy are run in the
context of the merged policy. The merged policy result is the sum of all
commands executed in the respective component policies CLI context in a
predetermined merge order.
If a merged policy already exists for a set of component policies, the merged
policy is used for the attachment instead of creating a new one. This
functionality allows a merged policy to have one or more attachments. A merge
policy is automatically deleted when the last reference is removed.
Chapter 2: Creating Policies
! 45
Merging Policies