sysmon

Usage

sysmon

Description

This is an extremely low-tech text-based multiple-system monitor running in a terminal window. The idea is to display some indication of health for a number of systems (specifically: 37 systems, although that changes from time to time) in a minimum of screen area. It also displays read-me notices for a number of mailboxes and Usenet news.

I'm not making any great claims to generality here. If there were 370 systems instead of 37, we'd have to think about something else. But (generally speaking) people tend to underestimate the amount of useful information that can be expressed in a simple terminal window.

The program runs continuously. Monitored systems are obtained from the file $HOME/.sysmonlist. Every so often, ssh is used to a remote system to run the uptime command; the results (assuming there are any) are parsed and displayed. Specifically, in the current implementation, the number of logged-in users, the system's uptime in days, the 20-minute load average, and the time it took for the system to respond are displayed. This is used as a semi-reliable proxy of overall system behavior.

Here's a screenshot:

There's a clock up in the corner, isn't that cool? Why? Well, because it fit.

Color is used tastefully to draw attention to unusual situations. Load averages over 1.0 are displayed in yellow, over 2.0 in red. Similarly, a response time over 15 seconds is displayed in yellow, over 20 seconds in red. Unresponsive systems get a big fat NO RESPONSE in eye-catching black-on-red.

The script also monitors multiple mailboxes and Usenet news semi-intelligently on lines 2 and 3 of the screen:

• Normal mail goes into $HOME/.mailspool/$USER on our system; the script displays the string [MAIL] (in red) if this file has a non-zero size.
• I use procmail to divert a number of incoming messages to mailboxes of the format $HOME/Mail/IN.name; the script displays [name] (in yellow) for each non-empty mailbox of this form.
• I also use procmail to divert postmaster and spam mail to $HOME/Mail/Postmaster and $HOME/Mail/Spam respectively. These are almost always junk, and although I want to scan through them periodically, I don't want to be nagged about it whenever they're non-empty. So the script displays [Postmaster] or [Spam] (in green) if the corresponding mailbox is non-empty and hasn't been accessed for some period (two hours for Postmaster and four hours for Spam).
• Finally, [NEWS] is displayed (in magenta) if the $HOME/.jnewsrc file is over four hours old, which limits my Usenet reading to something sensible. (Well, actually probably it's still not sensible.)

Features:

• The Term::ANSIScreen module available from CPAN handles text positioning, color, and formatting.
• The Term::ReadKey module (which is either included in Perl, or available from CPAN, I forget) handles non-blocking terminal reads. This was implemented so that pressing the Q key will quit, and any other key will clear the screen (in case it gets confused).
• The script uses a priority queue implemented as a heap (in turn implemented as a Perl array-of-hashes) to keep track of what to do when. (The code is adapted from Robert Sedgewick's Algorithms.) As each check is done (and removed from the front of the priority queue), the next occurrence of the check is scheduled (and inserted into the priority queue). Mail/News checks happen about every minute. System checks for each system happen at 4-6 minute intervals.
• An alarm signal is used to make sure we don't hang up waiting for a system to respond to an ssh; we give up after 60 seconds.

Source

Back to Hacks

Last modified: June 20 2003 09:12 EDT