After a run of over 200 days of uptime (and 22 months of overall operation), cedar, my company's server (which hosts several dozen sites), started to crash every four or five days a few weeks ago. This week, after trying a few other things, I asked my data center to run diagnostics on the hardware. I suspected perhaps a RAM issue. We scheduled a four hour window tonight from 1900 to 2300 CST for a motherboard and RAM test. If that didn't find anything, I had tentatively scheduled a 600 to 1000 CST slot on Monday to run a hard disk diagnostic.

As it turned out, the data center wrote me less than an hour after they started diags to tell me they had found faulty RAM, had replaced it and re-run the diagnostic successfully. Whether this will solve the problem once and for all, I'm not sure, but it certainly sounds good to me.