Details
-
Feature Request
-
Resolution: Unresolved
-
L3 - Default
-
None
-
None
Description
User Story (Required on creation):
In high troughput scenarios, job acquisition can introduce delay into process instance execution times and can create high load on the database. As an optimization, the cycle via job acquisition can be skipped by immediately submitting a job to the current engine's job executor. Note that this also circumvents the load balancing effects of multiple job executors in a cluster (e.g. all jobs may be created on the same engine).
Functional Requirements (Required before implementation):
- There is a configuration flag to enable that jobs on creation are immediately submitted to the local job executor; the job is immediately locked in that case
- If the job is rejected (queue full), then the job is unlocked (=> so that another job executor can acquire it)
- The behavior and effects of this setting are clearly documented and described as a feature for advanced use cases
Technical Requirements (Required before implementation):
Limitations of Scope (Optional):
Hints (Optional):
- A proof of concept implementation can be found in https://github.com/ThorbenLindhauer/camunda-engine-unittest/tree/job-executor-plugin