Keeping watch on 3000 hosts
October 25, 2006
HP 3000 servers may not need the babysitting that other environments demand. But when a server is mission-critical, as so many 3000s are, monitoring that resource's availability makes good IT sense.
New solutions for this kind of watching are rare on the 3000 market these days, but the existing product selection works just fine for HP 3000s — as well as HP-UX servers, Linux boxes, and even Digital VAX systems.
The software comes from ASP Technologies, based in Windsor, Colo. and it's been available for years now. Vantage is a console management package designed differently than many network watchers. This solution does not require installation of agent software on any managed resource like a 3000 or a 9000. ASP says this design "allows control over a much wider variety of systems and devices."
As a result of the Vantage design, no processing overhead is imposed on resources under Vantage control. All event detection and automation is performed 'outboard' on the management workstation.
Donna Hofmeister testified to the advantages of Vantage in a recent report to the user community over the 3000 newsgroup. In typical lower-case Donna-speak, she posted
vantage will watch your console traffic and will react according to what you've instructed it to do. the company owner (allen) knows mpe and is great to work with. the software itself is well written and well behaved. i think it's reasonably priced as well.
Veteran 3000 manager Greg Stigers added than many an SNMP-driven solution could fill the bill of requirements requested by Wesley Setree, who needs "a tool that will monitor HP 3000 and OpenVMS for certain conditions and send an alert via e-mail or pager and/or possibly a console or command center ... I would not only want to monitor down conditions but other items like disc space thresholds and job aborts." Stigers noted
While I’ve read that SNMP is becoming less important for system health and notification, it’s better than nothing. There are a number of tools that will query systems and handle SNMP responses. My last 3000 shop used What’s Up, although it never did much work to have it talk to the 3000, unfortunately.
I like to ask, who watches the watchers? Can you imagine a reasonable scenario, such as losing power or network connectivity, that would mean that the monitoring either was down or could not report the problem? Sure, it’s unlikely, but it’s not impossible.
At the other end of the problem, I will admit to writing a job that was a wrapper for a set of command files that checked the system for certain states, and reported their exceptions. Things like background jobs not running, or the number of jobs waiting increasing over three samples. That’s fine for OS & application states, local to the system.
Tony Summers adds that for the 3000 site with DIY skills, tracking solutions may be more flexible and engage more reporting devices:
For job tracking, we have written our own COBOL program which is invoked by the magic of system logon UDCs whenever a job is streamed, logs on, reports an error or logs off. By carefully managing the status during each of these events, we are able to report to the operators any jobs that have failed or didn’t reach their normal !EOJ job card.
A separate monitor program (another COBOL program) simply keeps watch on the job tracking files sending out alerts via various means — to the console, to e-mail and by SMS to the operator’s (and my) mobile phone (cell-phone).
We use a product called Scrambler from UK-based Gainsborough Software to forward the SMS messages from a Windows server to our mobiles. A simple FTP script transfers the SMS files created on the HP 3000 to the Windows server running the scrambler software.
As for monitoring disk usage — we currently have a job (run once per business day) that essentially inspects the results of a DISCFREE and send a similar email/SMS when the disc falls below a certain percentage.
I think the Scrambler product has built in functions to monitor the HP 3000 on your behalf, but our operators have never bothered to implement them.
Our monitor program also checks for other unusual conditions — for example, reporting when the global job limit has been set to zero.
I’m not suggesting you write the whole solution yourself, but given the limited market for the HP 3000 you might find you need to consider a bit of Do It Yourself.