Tinker Observability & error analysis

by Kondasamy Jayaraman · drawn with Colorfigures by April Zhu

Yesterday’s failed traces on the wall.

Chapter 2 of 7

The first failure

Some customers got wrong answers yesterday. Each bad reply has a trace behind it.

This chapter: Where, inside a bad trace, did things first go wrong?

You will learn

  • Why the reply is where you notice a failure, not where it starts
  • Finding the first span that went wrong
  • Why the first failure, not the last, is the one to note
Words you’ll meet (2)
First failure
The earliest span in a trace that did the wrong thing. Later spans often go wrong only because of it.
Upstream
Earlier in the trace. An upstream error flows into every step after it.

The picture on the left is live. Each step below asks you to do one thing in it; the story continues here as you go.

Step 1 of 5

Yesterday 18 traces ended in a wrong or useless reply. A customer saw each failure in the reply, but that is where it showed, not always where it started. Open the first.