Created an attachment (id=193)
patch for https://opennms.atlassian.net/browse/NMS-1259#icft=NMS-1259

Tarus,

Try this patch. I also fixed a few other things I encountered, such as 2 or 3
more SQL statements that needed ismanaged != 'D'.

No interface should be marked as secondary now unless it is in a
collectd-configuration.xml package.

--Bill--

> Actual data collection controlled by collectd should not be adversely affected.
> Have you noticed some undesirable collection behaviour as a result of this?

Yeah. We have a customer with lots and lots and lots of interfaces. Some of those are on VLANs and
they don't care about them. Since they use "select" as their collection parameter, secondary IPs
automagically get collected. When I excluded them from the collectd package (and thus removing them
from "secondary" status) they never changed to "N".

This is important because the system is collecting over 145K objects per poll, and the more we can
minimize, the better.

In the meantime, I don't think there is any harm in allowing the GUI to change the status from "S" to "N",
so I plan to update the code to allow this.

"Okay, for an interface to be a Secondary snmp interface, it has to qualify to
be a primary interface. This means it has to be in a valid collectd package"

The above statement is not necessarily true anymore. The code first tries to
determine a primary based on a valid entry in a collectd package, lowest
loopback first, then lowest address. If that fails, it looks for the lowest
loopback address without regard for collectd. If there are none, it picks the
lowest address, again without regard for collectd. All other ip addresses on the
node that support snmp are marked 'C'.

In practice 172.20.1.110 should not become primary unless there there are no
other valid ip addresses for this node in a collectd package.

Actual data collection controlled by collectd should not be adversely affected.
Have you noticed some undesirable collection behaviour as a result of this?

--Bill--