Database Emergency Response Plan: The 3 Cs
Preventing and responding to database performance problems is a critical component of our responsibilities, and we need to have a solid game plan to be effective. This document is a general guide to dealing with DB performance issues. It is a living document, and as we encounter and deal with emergency situations, we will continue to build up our toolbox and experience.
Steps
Notice the clues
The first step is always to notice the clues that point to a performance issue; this will help us determine the appropriate response to the problem. In some cases, multiple metrics show issues, so we need to build up a bank of responses based on combinations of clues and individual ones to understand which clues represent urgent, critical problems.
Diagnose the cause
Once we have the various clues to work from, the next step is to find the underlying issue causing the problem. None of the clues are directly fixable themselves, so we can only fix the issue once we find the cause.
Attempt a cure
The specific cure will depend heavily on the problem. Some will require a small response, while others might only be solved by restarting the database.