Many beginners “guess.” Real DBAs measure first.
Examples:
“Checkout page is slow”
“Report takes 3 minutes”
“Users can’t log in”
“Database is full”
Write the symptom clearly. Do not jump to the solution yet.
Evidence is objective. Examples:
Wait stats show I/O waits
Blocking chain shows one session blocking many
Query Store shows a plan regression after deployment
Error log shows repeated failures
Most production issues fall into these buckets:
CPU pressure (too many expensive queries)
I/O pressure (slow disk reads/writes)
Memory pressure (insufficient memory, spills, poor cache)
Blocking/deadlocks (concurrency issues)
Bad plan/regression (optimizer chose a worse plan)
Configuration (MAXDOP, memory settings, tempdb setup)
Safe first moves:
capture evidence
stop the bleeding (short-term workaround)
apply a targeted change with rollback
Unsafe moves:
“Let me rebuild everything”
“Let me change 10 settings at once”
“Let me restart production”
(These can cause outages.)
Validation must be measurable:
response time improved
CPU dropped
blocking reduced
error stopped repeating
no new errors introduced
Your documentation is part of the fix:
what happened
what evidence you saw
what you changed
why you changed it
how you validated
prevention plan
Bad approach:
“I’ll rebuild indexes and hope it fixes it.”
Professional approach:
Define: “App queries slow since 9:05 AM.”
Measure: “Many sessions waiting on locks.”
Isolate: “Blocking chain caused by one long transaction.”
Change: “Coordinate with app owner, end blocker, reduce transaction scope.”
Validate: “Response returns to normal, blocking clears.”
Document: “Incident summary + prevention steps.”
Try it now (10 minutes):
Write your own 6-step workflow in your DBA Notebook using a made-up incident:
“A report is slow”
“A restore is needed”
“A login fails”
Choose one and fill all 6 steps.
Not a member yet? Register now
Are you a member? Login now