Re: snmp in cluster

[Fulko H. <ful...@gm...>]

On Tue, Nov 1, 2011 at 9:09 PM, Nguyen Dinh Phong <nd...@ya...> wrote:
> Hi,
> I'd like to set up a cluster of nodes running net-snmp5.3.3.2, with only one
> main node responsible for external requests. The requirement is to make the
> whole cluster appearing as one system to NMS applications. If a customer
> does a snmpwalk at the main node, he would see all the OIDs for all the
> nodes in the cluster. For example with the system mib:
> system.sysdescr.0 for the main node
> system.sysdescr.[index1] for node 1
> system.sysdescr.[index2] for node2,
> ...
> With all nodes running the same MIBs, is agentx suitable here? How'd I
> choose a scheme for indexing different nodes of the same OID?

When I was forced to do this, I invented my own MIB that 'tableified' all of the
information that I needed to aggregate across the 'cluster'.

I then implemented that MIB with software that interpreted the indexes to find
out what 'device in the cluster' I needed to access, and then (as a proxy)
retrieved the remote data, and responded (inside my 'aggregate' MIB).

So as the other response said, "you can't add indexes to an existing MIB",
but I created my own MIB that had indexes.  I.e.

...myMib.sysdescr.[index]