- Changing how DAG is represented, both in code and how DAGs are defined in JSON. - Removing std::vector<Task> representation in favour of a map that will enforce unique task names - Task names now have a name (generated), and a definedName. - Adding support to loggers to add tasks after a DAGRun has been initialized.
1.3 KiB
1.3 KiB
Tasks
-
Open
- REST Server
- Add in authorization scheme (maybe PAM auth endpoint with JWT?)
- Flesh out server and interface
- Core Functionality
- Allow
{{,!DATES}}style interpolation, where values from{{DATES}}are preserved as a single, comma separated list - Allow for tasks to define next tasks
- Refactor [de]serialization so that a task can be parsed by itself
- Add notation of parameterValues
- Tasks are now refered by two names:
- baseName is the original name in the spec
- name is the individual tasks
- Add execution gates
- Executors
- Slurm Executor
- Loggers
- FileSystemLogger
- Add unit tests
- Add more error checking
- General logger
- Redis DAGRunLogger
- FileSystemLogger
- Allow
- REST Server
-
Completed
- Core Functionality
- Add ability to define child -> parent relationships
- Rip apart scheduler and re-implement runDag as a utility function, taking an execution pool, task list, and logger
- Resume a failed DAG
- Handle return on errored DAG / Task
- Clearing a DAG Task
- Core Functionality