I was extremely unkind to my new computers, but apparently they’ve forgiven me.
What happened was, I launched the first really big Zeppelin test run on my two V20z boxes, and I was watching the diagnostics and realized that I’d got one piece of the design really wrong; all of a sudden there was a nice refactoring shrieking at me to be done. So I dove back into NetBeans on my Mac to do that, but I decided I might as leave the test running, the results might be interesting.
Well, most of a week passed while I got distracted with Project Coyote and Atom and a couple of other things and did the refactoring in nooks and crannies of time. This afternoon, I logged back to the V20z’s and poked around and all my JVM’s still seemed to be there, which was a bit surprising since I’d miscalculated the memory and expected them to blow up, eventually.
So I probed the Zeppelin and the probe locked up and I decided something was wrong and eventually killed the login, only then I couldn’t log in again. Eventually, after a few minutes and a few tries, I did get back in, and foolishly ran the probe again and it came back instantly; the Zeps were actually still aloft.
To make a long story short, due to a slight miscalculation I had JVM’s running substantially exceeding the physical memory, plus due to another, the logfiles had filled the disks up to 100%. And near as I can tell, the program had gone on running, trying to use more memory and write more logfiles, for the six days I’d been ignoring those computers. I’m kinda surprised there wasn’t smoke coming out of that corner of the lab.
Anyhow, I nuked the JVMs and wiped the logfiles and both systems are back to normal, near as I can tell, uptime of 20 days and happily running more Zep tests. The conclusion is obvious; for standing up to egregious abuse by overenthusiastic ham-fisted programmers, Solaris 10 is da bomb. Not too bad for an OS that just shipped.