Here is a graph of swap usage on an IPC running solaris 2.5, showing a couple of things. First of all, there is the steady growth of unavailable swap (the orange data). This graph shows roughly one week of up time, but eventually the swap growth makes the machine unusable. Second, note the pretty obvious correlation between the swap growth and the virtual size of cron (the red data) over time. Frequently, these cron jobs dump core after a machine has been up for a while, often making the /var filesystem unusable.
The other data sets show that xrn, netscape, xmh, and several big pieces of CDE do not appear to be responsible for the growth in swap usage. They occasionally go up or down in virtual size, but don't exhibit the steady growth that cron does.
The blip at the beginning shows a short period of time during which CDE and its ancillary procedures were not running at all. After that there is a reboot and CDE was started. Note that the rate of growth of both cron and swap usage is the same whether or not CDE is running.
click here to view the gif version of the graph
click here for the postscript version
