detective work
Posted in 2000
Thanks for everyone's suggestions about our Informix troubles of last week. Here's what we ultimately found out. Strangely enough, our various lock-ups were ultimately traced to a malfunctioning cooling fan on the physical machine. Intermittent overheating caused the physical machine to shed processes to keep from burning up, which in turn involved killing Informix back-end processes. We'll probably never be 100% sure about this, but the correlation in terms of when the problems ocurred is so strong that I'm reasonably sure that it was not a coincidence. The key to uncovering the link was reading through all the unix SYSLOG entries during the interval when Informix was misbehaving and looking for anything unusual. Once I discovered the temperature warnings, comparing greps of the online.log with those in the SYSLOG quickly established the three-way correlation between Informix problems, temperature warnings, and killed unix processes. Thanks again. And fingers crossed. I don't know if I could take another week like last week. Mark McDonough Knowledge Systems Inc.