See the reasoning behind a finding.
Two reproduced PostgreSQL problems with captured plans, locks, timings and tested responses.
Investigation examples
Measured lab
Generic plan regression
A plan setting hurt small tenants. Follow the change, plan evidence and tested response.
Measured labDDL lock queue
An idle transaction, one ALTER TABLE and 14 seconds of zero reads, reproduced and measured.
These are synthetic labs using generated data in an isolated environment, not client incidents.
What a lab case shows
Each lab reproduces a kind of production problem I have worked on, using a fresh PostgreSQL instance and generated data. The setup and workload scripts and captured output are public, so you can rerun them and check the numbers. The point is the reasoning: which evidence pointed to the cause, which response was tested, and how the result was validated. That is the same structure a client report follows; see the sample report.
Client cases
Client engagements are confidential by default. A client case appears here only after the client has approved an anonymised write-up in writing. None is published yet.