Why do you want to contribute?
I am a heavy cloud and selfhost trigger dev user.
I have 5 years of dev experience. And I am looking to work on this feature:
Allow a task definition to declare the region(s) it may execute in:
export const myTask = task({
id: "my-task",
region: "eu-central-1", // pin to one region
// or
region: ["eu-central-1", "us-east-1"], // allow one or more regions
run: async (payload) => { ... },
});
The use case for this is:
Resilience. If one region is down or at full capacity, another listed region can still pick up the run. Since workers pull, whichever worker in any listed region grabs it first runs it.
Checkpointing on mixed CPU architectures. With ARM and AMD machines, a checkpointed run can't be restored on a different architecture (CRIU), so those tasks need to be pinned to one architecture. That means splitting the workers into an ARM group and an AMD group. The downside of the split is that a general task pinned to one group only gets that half of the fleet. Listing both groups on the general tasks gives them the whole fleet back, while the checkpoint-sensitive tasks stay pinned to one.
Prior contributions or relevant experience
No response
Why do you want to contribute?
I am a heavy cloud and selfhost trigger dev user.
I have 5 years of dev experience. And I am looking to work on this feature:
Allow a task definition to declare the region(s) it may execute in:
The use case for this is:
Resilience. If one region is down or at full capacity, another listed region can still pick up the run. Since workers pull, whichever worker in any listed region grabs it first runs it.
Checkpointing on mixed CPU architectures. With ARM and AMD machines, a checkpointed run can't be restored on a different architecture (CRIU), so those tasks need to be pinned to one architecture. That means splitting the workers into an ARM group and an AMD group. The downside of the split is that a general task pinned to one group only gets that half of the fleet. Listing both groups on the general tasks gives them the whole fleet back, while the checkpoint-sensitive tasks stay pinned to one.
Prior contributions or relevant experience
No response