888

IBM RS/6000 systems have a three-digit LED display on the front for diagnostic purposes. An AIX priest will have RTFM, will be familiar with many codes that can appear there, and can tell you exactly what the machine is doing as it boots up.

IBM makes some parallel-computing supercomputer clusters, called the SP/2, made up of several RS/6000 systems. The control console comes with an XWindows application that displays the LEDs of the individual nodes!

The point of the story is, when an RS/6000 thoroughly wedges itself, the display flashes "888" as a sign that something is severely fucked, and more diagnostic codes are available for retrieval. Thus amongst AIX sysadmins, 888 has reclaimed its association with the dark and sinister.

On (older) rs/6000 CPUs what the diagnostic LED's* would flash when the kernel panics.

Flashing 888 on these AIX systems are a clear indication that the machine has crashed and needs to be rebooted. Post-crash frobbing of the reset button will cycle the diagnostic processor display through a series of codes. These codes indicate the nature of the failure (software, hardware and possibly which subsystem crashed), as well as indicating the status of the system dump* (in progress, complete, partial).

---

Notes:

A less well known use of the LED on rs/6000 / AIX systems is to run a daemon known as loadavgLED, which takes the system load average (getloadavg(3)) from the kernel and displays it to the diagnostic LED. Running this hack is nearly certain to get the attention of any IBM rep, as normally it's only lit up during bootup or something bad has happened.

Taking a system dump is the process of saving an image of virtual memory, processor status etc for later forensic analysis. The AIX system dump documentation (entitled "How to take a system dump") is quite amusing reading, (possibly inspired by the context that tech support needs to convey this faq to customers in a state of panic). See also deadbeef.