hpasm and snmpd traps - how to test? how to configure?

[Raphae]

Normally snmpd.conf is configured with directives to monitor certain conditions and to take an action if a certain threshhold is exceeded. I can find no information anywhere as to how the various hpasm components are configured to send snmp trap events for specific conditions. I am primarily interested in having trap events for excessive temperature readings. I can use "hpasmcli -s 'show temp'" and see the readings, and I have cacti configured to collected the sensor temperature data and graph it, but my understanding is that hpasm should also be able to send snmp trap events.

There seems to be a dearth of information about the actual implementatin of hpasm - perhaps because it is a pre-made Redhat package designed for people who only want to click on web interfaces and not really actually know how snmp works. At any rate, I managed to build a Debian package from the RPM and install it on several Proliant DL380 G4 servers, no thanks to HP for creating a Debian package or providing hapsm in some format that conforms to standardized ways Open-Source projects are packaged.

Finally, there is no information on how to test the snmp portion of hpasm. It would be useful to be able to simulate events to see if the traps are actually working. Isn't this the kind of thing that most sysadmins would naturally think of doing?

[David Claypool]

The 'hpasm' package interfaces with the Advanced Systems Management Controller. See http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&objectID=PSD_CN0416W for more information on it.

The easiest way to 'test' for most people is to pull a drive in the array.

[Raphae]

Does hpasm or does it not send snmp trap events for things like critical temperatures, and if so, where is it configured? My understanding was that it is capable of doing this...

[David Claypool]

"Power, Thermal Events, Memory & Processor ECC

HP ProLiant servers have special management hardware that allows the HP ASM software to monitor Power, Thermal, Correctable Memory and Correctable Processor events. ***The HP ASM software will note any changes in the system and will notify the HP System Insight Manager (SIM) and Insight Manager 7 (IM7) of the event***. The HP ASM software understands the thermal operational policy for each individual server and will initiate a graceful OS shutdown should the thermal parameters exceed the thresholds set by the policy. There are many variables that factor into the decision to initiate a graceful OS shutdown. These variables include the presence of redundant fans and power supplies and the operating environment factors such as temperature.

The HP ASM driver will monitor correctable memory and processor errors. While some correctable errors are expected during normal operation, an excessive amount is an indicator of imminent failure. An error will be logged to the HP IML log and notification sent to the Insight Manager console via the HP Insight Management Agents executing on the ProLiant server."

[Raphae]

I'm sorry but these canned replies are not even addressing my questions. I appreciate your trying to help, but unfortunately you do not seem to have enough experience actually configuring SNMP in production systems to understand my questions.

Here is an excerpt from an HP doc regarding the Threshold Agent, which runs as a process cmathreshd:

"The Threshold Agent implements the Threshold MIB. Users can set thresholds on counter- or gauge-
type MIB variables. The Threshold Agent periodically samples each selected MIB variable at a rate
defined by the user.
MIB data values are compared to user-configured thresholds. If a configured threshold is exceeded,
an alarm trap is sent to the configured SNMP trap destination and to Linux email (configurable
through trapemail entries in /opt/compaq/cma.conf file). User-configured alarm thresholds are
permanently saved in the data registry until deleted by the user."

If you understood SNMP, you would know what an MIB and a trap is. Given that it explicity states here in this HP document ("Managing ProLiant servers with Linux") it would be helpful to know precisely how these thresholds are configurued (i.e. what specific configuration file) and also where the trap host is configured (I assume it is in /etc/snmp/snmpd.conf since there hpasm did modify it upon installation. So at least we know that hpasm tries at least minimally to cooperate with the existing ucd-net snmp infrastructure rather than imposing its own hacked version over it.)

[Rob Buxton]

When you installed the HPASM Package it would have asked questions on configuring SNMP, that includes Trap destinations, host access, community names etc. They're typical snmp configuation requirements

Yep, they're in the snmpd.conf file, well they are on my VMWare ESX Server which is a Linux variant.

As for the thresholds, on Windows the thresholds are, as David details, unchangeable except for items like Disk Space etc.

I don't believe they'd be changeable in a Linux environment as I do not think HP would want thresholds governing temperatures etc. changed.
The configuration of those items is nothing to do with SNMP, SNMP is just the messenger boy.

[Raphae]

Re: threshholds I see a couple of problems with this. First, with normal SNMP directives you can specify how that directive behaves. "man 5 snmpd.conf" and see for example the section on the proc directive. So in the case of hpasm we have what apparently are things that are supposed to behave like normal SNMP directives but aren't *really* directives?
Also, if one cannot set the threshholds or even find good documentation on what the values are supposed to be, how can one rely upon hpasm for critical situations? Before I deploy it and consider it "production" I would definitely want to set the threshholds very low to intentionally trigger alerts to verify that the notification/shutdown processes work. I would need to do extensive testing of this system before I would consider it reliable enough to deploy in a critical environment.

Re: the SNMP setup: yes it did in fact prompt me for a bunch of questions when it installed - but this was not really necessary since I already had an existing SNMP setup...

[Pedro VERA DE LA ROCHA]

Hi, Raphae.

I know this thread is a little old but I am really interested on knowing how did you solve your questions.

I am trying to manage several Proliant systems through SNMP by using HP Server Management Agents, but I cannot make it work.

I would like to provoke some traps by changing a disk threshold or shutting down a NIC interface, but no success, though I think I have correctly configured snmpd.conf.

The documentation I have found looks very schematic, and it seems you need HP SIM in the manager node to be able to work, but I am using instead a generic SNMP manager.

If you could give me some hints, I would very much appreciate.

[Claudio T. S.]

Try
/etc/iniit.d/hpasm reconfigure

Not sure about , but in the old version , when you're installing the agent , the configuration ask you about the SNMP configuration , but in the new version default configuration is saved.
With hpasm reconfigure , the system will stop the agents and will ask you about SNMP configurations .
To test , I put my Cisco switch in shutdown and I was able to receive the trap.
Hope this will help.