- High: It blocks me to complete my task.
Is there a built-in watcher/API to kill an actor/task if it’s heap memory usage (or another custom resource) is above a certain threshold? If not, is there a best practice/guideline to implement such a mechanism? Is it a good pattern or an anti-pattern?
Ray should automatically kill workers if the heap memory usage on the machine overall grows too high, and you should get a scary error message when this happens.
Recently, we also added a memory monitor that gives you a bit more control over this behavior. You can check it out here and please file an issue on github if you run into trouble or have feature requests.