Let Open Source Open 3000 Doors
September 28, 2010
A whole class of open source software -- tools and utilities -- can or can be made to work with the HP 3000, even if the software runs on other equipment. That's the advice from Brian Edminster, who owns Applied Technologies, a firm consulting to and managing HP 3000 sites both homesteading and in transition.
For example, Edminster says, the network graphing and trending tool Cacti can be feed by SNMP. "The 3000 has SNMP interfaces for quite a number of key statistics," Edminster said. Nagios is another tool that just needs a data-collection client to run on the 3000, allowing an MPE/iX box to be monitored just like any other piece of equipment in a company's network.
One of Edminster's customers is using something similar to but simpler than Nagios: Xymon (formerly known as hobit) to monitor and trend server statistics, as well as issue alert notices when user defined thresholds are met. "I've got a prototype client for the 3000s here, so they can be monitored and alerts generated," Edminster said, "just the same as any of the several hundred other servers in this network. And yes, I'll be making the Xymon client for the 3000 available as soon as it's fairly stable; it's still being tweaked."He says these free tools are a pretty good example showing how open source tools can be used to help a 3000 fit in with existing infrastructure, and draw benefit from those tools as well.
Solutions from open source such as Nagios need a data collection client to operate native on the 3000, and "I don't know of any third party collection clients for MPEi/X -- but that wouldn't preclude someone from writing and even selling one," Edminster said. "I'm also not aware of any open source collection clients for MPEi/X, but I'd expect that the clients available for various versions of HP-UX would be a good starting point if someone wanted to make one."
Nagios can be configured to monitor 'publicly available services' such as FTP, telnet or http right out of the box,' even when those services are provided by a 3000. "Anything else requires one sort of collection client or another," Edminster said. "There are both passive and remote invoked collection models in the Nagios design, and either could be made to work."
The development work to throw open the open source doors "isn't really rocket science, depending upon how deep the data you're seeking lies inside MPE."
Most of the time, Edminster explains, 3000 managers are interested in fairly pedestrian things, like user count versus limits, job counts, number of waiting jobs, free space, or service availability. All of which can be done either by native capabilities of existing monitoring tools, or via simple shell or CI scripting under MPE. It's even possible to get things like average CPU via use of SNMP calls or CI commands.
"It would be possible to write something more 'run-time' efficient if you had an MPE internals expert reading the internal data structures directly," he said, "versus using the 3000's Measurement Interface, which is notoriously inefficient. "But in most cases, that's really not necessary. I prefer to use the simplest, easiest to maintain method to get data. If that's not fast enough, then we get more exotic -- and only if absolutely necessary."