Skip to content

20170831_srecon17_1.5

20170831_srecon17_1.5

Debug after bad deployments

Intro

Context

Bad deploy happens
* Many commits
* Many deploys happens, which was the cause ?

Basic resolution
  1. Rollback
  2. Debug
    + Find the errors
  3. Fix
  4. Restart
Main types of errors
  • Perf regressions
  • Broken pages
  • Empty results
  • Unexpected status code

Debugging

Find the NEW stack trace
Find the bad commit

1 - Find the errors
  • Logs aggregation
  • Identifying NEW stack traces
    -> Compare 2 ES Queries by example
    -> Deduplicat similar stack traces
2 - Find the bad commit

Aka: Wich of the 71 commits did that ?

https://github.com/pinterest/git-stacktrace

git stacktrace COMMIT_ID..COMMIT_ID < STACKTRACE