lbilger/camunda-deadlock-reproducer

Deadlock reproducer

This is a really minimal process application that can be used to reproduce the deadlock issue.

It is only meant for local startup.

The code does not show any best practices etc., it is meant to be small, quick and dirty, not clean and maintainable!

Running the engine

• Start the required containers using docker-compose:

  docker-compose up -d

• Build with maven:

  mvn clean install

• Run the application

  java -jar target/camunda-deadlock-reproducer-1.0-SNAPSHOT.jar

Using the engine

Now you can:

• start 1000 new processes:

  for i in {1..1000}; do echo -n "$i "; curl -X POST localhost:8082/public/process-rest/simple-process/start; done
  # or using parallel
  parallel -l 1 -j 10 'echo -n "{} "; curl -s -X POST localhost:8082/public/process-rest/simple-process/start' ::: {1..1000}

• complete all tasks:

  curl localhost:8082/rest/task | jq -r '.[].id' | while read taskId; do curl -X POST "localhost:8082/public/process-rest/simple-process/task/$taskId/complete"; done
  # or using parallel
  curl localhost:8082/rest/task | jq -r '.[].id' | parallel -l 1 -j 10 'echo -n "{#} "; curl -s -X POST "localhost:8082/public/process-rest/simple-process/task/{}/complete"'

• start a batch to delete all historic process instances:

  curl -X POST localhost:8082/rest/history/process-instance/delete -H 'Content-Type: application/json' -d '{"historicProcessInstanceQuery": {}}'

You should now see some deadlock exceptions in the log. If not, repeat the above, maybe with even more process instances.