When things go wrong or begin to look strange to an experienced eye, it is time to investigate. But where do we start?
Is the redundant array of independent disks (RAID) running in parity mode, thereby drastically reducing the I/O throughput? Is the upstream switch saturated, robbing the database of bandwidth? Are we out of memory and swapping to disk, or are we causing memory reclamation threads to terminate processes? Has the operating system task scheduler gotten overloaded and spiraled into oblivion?
Maybe! We've seen all of these scenarios, and many more. We can't fix a problem that we are unable to locate. Any time that we spend analyzing an unlikely path is ultimately wasted, and it only increases downtime. We must take an inventory of the known symptoms and extrapolate this evidence into one or more avenues of investigation.
Anything less is...
