Setting up Managed Systems in ICE-Linux without the ProLiant Service Pack

However, HP SIM can be configured to respond to an SNMP event (or trap) of type ColdStart to

automatically discover the server associated with the IP address in the received trap. Installing the

ICE-Linux plug-in automatically enables the discover on event feature of HP SIM.

On receipt of a trap, if the server with the included IP is recognized as one already in the HP SIM

database, the complete discovery procedure (that is, discover, identification, and data collection) is

abridged to just discovery. If the server on that IP is not recognized, SIM performs a full discovery,

identification, and data collection set of procedures, and a new node is created in the database.

If HP SIM has a previous record of a different server in its database for that IP address, it will

automatically orphanize the old node to preserve any historical information in the database for that

node, and to create a new node in its database for the newly discovered node.

Additional event handling configuration options can create specters, that is, nodes waiting to be

born. One such option is the “create a node when management processor discovered” option.

Configuration and Data Collection

After discovery, HP SIM normally checks the health and status of the hardware and the software of the

server at regular intervals and updates the icons in the collections display. Status icons are a

combination of health and event (trap) status. Clearing all events does not necessarily return the

health status to green until the next status poll.

Independent of status polls, tasks are scheduled to be run automatically to perform regular data

collection on managed nodes. These tasks collect, among other things, the MAC or Media Access

Controller hardware addresses for the Ethernet Network Interface Controllers (NICs or Ports).

Unfortunately, as part of the data collection procedure, the task removes any existing information,

including the MAC information, from the HP SIM table for the node. Thus, it is possible for a node to

loose its MAC information while still appearing to be manageable if the server is heavily loaded and

the SNMP agents cannot respond within the timeout window designated for the SNMP data collection

event.

SNMP

SNMP agents can be considered as web servers with web pages to be served. The client (in this

case, the CMS) uses an SNMP browser tool like snmpwalk to read or walk those pages. Each page

has a URL, or in the case of SNMP, an Object Identifier address (OID), to fetch the information from a

particular page. One merely needs to know its address, the OID.

SNMP also has the equivalent of a static name service called a Management Information Database,

or MiB. MiBs are handy for translating an OID numbered address into human-readable form. They

are, however, merely for convenience and must be manually installed on the client performing the

query to translate human-readable addresses into OIDs and to translate OIDs into human-readable

form. They are not generally required to use SNMP.

The OID address space is treated like an inverted tree and branches into different information zones.

Thus, most of the address for a particular zone resembles the address for neighboring or adjacent

information.

Branches of the tree which are serviced or actually return information are provided with the

information to be served either by the original SNMP master agent or by subagents that have come