Small caveat: I spawn snapshot_manifest on search runtime (cause it's a read-only op on segments) and restore_shard_snapshot on update runtime (cause, well, it's defo not "search"). Given the issues we've seen, I'd say maybe possibly this might have a small potential for a dead/live lock. Maybe.

But I'm also not sure if it would be any better to spawn both tasks on the same runtime.

But I'm also not sure if it would be any better to spawn both tasks on the same runtime

I think it should be better. Like deadlocks happens when one thread is waiting on other to free up stuff. But if it's all in the same runtime/thread, deadlock shouldn't happen, right?

But I'm also not sure if it would be any better to spawn both tasks on the same runtime

I think it should be better. Like deadlocks happens when one thread is waiting on other to free up stuff. But if it's all in the same runtime/thread, deadlock shouldn't happen, right?

In this case there's a spawn and a blocking spawn. Even within the same runtime they would already be isolated in different 'pools'.