Introducing the new Sun x86 SNMP Hardware Monitoring Agent

27 May 2009

I thought now it would be a good opportunity to introduce
the newly released Sun x86 SNMP Hardware Monitoring Agent found in the Sun Server Hardware Management Pack 1.0 to discuss its functions and high-level architecture. First what is the function of this agent? Basically
it is a host-based (i.e. in-band) SNMP hardware monitoring agent for Sun x86 servers. In the 1.0 release, the Sun Server Hardware Management Pack can provide:

Monitoring of environmental sensors
Inventory and presence of FRUs
Status of system indicators
Trap notifications for all sensor-related events

This enables Sun x86 systems to provide SNMP monitoring from the host OS
to anything that can consume SNMP which is basically the vast majority of
the Monitoring Systems out there in use. This agent is not yet utilized by xVM OC (Ops Center) as that is already highly integrated with Sun platforms via both IPMI and SSH and can easily monitor, control, and provision Sun ILOM (and also ELOM, ALOM and RSC) based systems.

Platform Support: Current Sun x86 ILOM server platforms. Since this agent installs within
the host Operating system it does not support the Blade Chassis
Management Modules (CMM):

Solaris x86
Sun supported Red Hat and Novell Linux distributions
VMware ESX 3.5
Microsoft Server 2003 (Q3CY09)
Microsoft Server 2008 (Q3CY09)

Architectural Overview

This new agent provides in-band monitoring to complement the existing out-of-Band monitoring
provided by ILOM itself. This is really a question of choice rather than there being a
correct approach, some users like out-of-band monitoring
over private management LANs direct to ILOM whilst others prefer
in-band communication with agent(s) in the host OS. We know that a significant proportion of users do not connect
Service Processors to the network. In addition users may already utilize
other monitoring agents on the host OS so it may be the preferred point
of monitoring.

There are 2 fundamental components plus the KCS driver that provides a
path to ILOM from the host OS:, a daemon hwagentd and the SNMP agent itself.

x64 agent architecture.png

First the hwagentd daemon’s role: As illustrated in the figure above. hwagentd polls
ILOM via IPMI over the KCS driver for events that it then
communicates to the SNMP agent (and potentially any other future
agents for example WS-MAN). In addition it can write events to syslog. The hwagentd
polls every 30 seconds and caches the data.
In the future hwagentd due
to its architecture can also call other libraries/utilities, thus extending both it’s own functionality and as well as the functionality of ILOM.

The agent supports a pair of MIBs (Management Information Base) SUN-HW-MONITORING-MIB defines the SNMP GET interface and SUN-HW-TRAP-MIB defines the SNMP Traps (event / alert) generated by the agent for net-snmp and can both receive and request events from the daemon. The agent does not itself communicate directly with ILOM but rather makes calls via hwagentapi and reads the hwagentd
cache. It can then propagate these events to an external management platform or tool or respond to SNMP trap requests. Key features of the agent include:

SNMP Interface – Obviously! Still the de facto system/network management protocol
Support for AgentX – Allows for flexible configuration of the OS agents and improved co-existence with other agents.
C API Interface – This will allow for ISV
integrations to leverage the polling and aggregation logic while not
assuming the SNMP overhead.
Trap interface – This allows for event-based integrations with ISV applications.
Identical sensor names as the SP – Consistent naming not only presents a more unified product but also improves serviceability
Logical grouping of information in MIB by device type – The
information in the MIB will be presented in logical groups that present
an organized view of the hardware system.
Aggregated status information – The ability to have rolled-up
status for all sensors and groups of sensors by device type will allow
ISV integrations and users to quickly detect if there is an error in
the system.
LED State - state of system LEDs visible

What about ILOM based SPARC Platforms?

This agent complements what we already offer
for SPARC, specifically MASF (Management Agent for Sun
Fire). The MASF agent is an SNMP agent that runs on Solaris SPARC and
exposes the SUN-PLATFORM-MIB. On ILOM-based platforms it communicates
with ILOM by a somewhat different architecture than that shown above.
It has extensive SPARC platform support including: T Series (including
Uniboard), V125 to V445, Netra 210, 240, 440. Details here:

Download Sun SNMP Management Agent 1.6
Documentation

What Next?

Most immediate on the roadmap is Sun Server Hardware Management Pack 1.1:

ILOM 3.0 Support
Red Hat EL 4.0 Support

Then in the near term, Q3CY09, Windows Server support is the next big
add. Also we will see increasing support of the agentry within our own 3rd party System Management Integrations. Over time we will release a new WS-MAN agent enabling us to support CIM (Common Information Model), which is seeing growing support particularly in the Windows and VMware world.

Where Can I Learn More?