Blog Rss Feed
At work, I manage a cluster of several hundred nodes. All those nodes are equipped with a little LCD display, like to one showed in the picture on the right. Until now, these displays were only displaying the name of the cluster vendor at boot, and the node name when the machine was up an running. To put these neat little displays to better use, I wrote a Python program named lcd.py, which prints more useful information.
When started, lcd.py burns into the MVRAM of the display the name of the company I am working for and the name of the node, which is in fact the position of the node in the rack. This helps me find which node is which when I take several of them out of the rack for maintenance. I just power the machine for a few seconds to read its name and position on the racks on the LCD display .
Then, the daemon goes into an infinite loop to probe the disk temperature using the S.M.A.R.T information provided by the disk. This information is then displayed on the second line of the LCD display. A quick look at the display and I know if something is wrong with the disk or the machine. Of course this is not the only monitoring system used to watch the system but every little detail helps. And it is fun to use these little displays.
The program also pings the “master node” to see if there are any network problems. If the “master node” cannot be reached, the program turns off the LCD display backlight. A quick glance at the entire rack and you know which node is in trouble.
For more information on these LCD display you can download the manual.