* Include shared state to allow graceful shutdown
Graceful shutdown of executors allows the current task to finish before
exiting. This prevents half completed tasks in the general case: when
workers are being scaled down.
To accomplish this a shared state (using an `Arc<RwLock<WorkerState>>`)
is created in a WorkerPool on instantiation. This shared state is then
passed to each thread (spawned with `WorkerThread::spawn_in_pool`), and
finally passed to the `Executor` instantiated by the `WorkerThread`.
This allows the infinit loop in the executor to receive signals from the
`WorkerPool`, and exit gracefully when requested.
* Add basic error handling
Add `FangError` enum derived from `thiserror::Error`. This should be the
default the error type for the Fang library, all errors returned by Fang
should be a value in this enum.
Use FangError for errors returned by start and shutdown, remove unwraps.
* Include instructions for running tests locally
* Track handles of panic'd worker threads
Allows the Drop trait impl of ThreadWorker to access the
`thread_join_handles` of WorkerPool so it can update the thread handle
when the previous thread unexpectedly exited and a new one is being
started. This is done in a way that prevents data leaks (by using a
Hashmap keyed off the name of the worker thread). It also ensures that
threads started from the Drop impl are properly joined on shutdown.
* Fix WorkerThread drop implementation
WorkerThread can not have `Clone` derived on it, as each cloned copy
will try to restart the thread when it's dropped, leading to an infinite
number of thread spawns till stack overflow. Oops
* Remove Option from SharedState type declaration
Instead of having an Option wrapping an enum, have the option codified
as a state in the enum.
* Bump version to 0.5.0
* Add integration test for shutdown
* Update simple_worker example to include signal monitoring and shutdown
* Update readme to mention using signal-hook to gracefully shutdown worker
* pass PgConnection into run function
Changes:
- rename Postgres into Queue
- pass PgConnection into run function
* remove all tasks of specific type
* add pgconnection
* pass connection to every queue function
* add connection pool
* reuse connection in test
* update readme
* Periodic tasks
* schedule next execution
* add scheduler
* ignore test
* fix clippy
* make start public
* check if the period task already exists
* do not insert task if it's already in the queue
* fix tests
* execute different types of task in separate workers
* add more tests
* remove finished tasks
* add retention_mode
* make db url optional
* Add worker params
* Add CHANGELOG entry
Rafael Caricio
