How do I handle a cpu intensive task at the start of the program without blocking other http probes?

[persona94]

My flask app runs on gunicorn that has 1 worker and 8 threads. At startup the application reads a few million lines of JSON to load into an in-memory database. To do this, I spawn a gevent that does the read. I yield several times (I use gevent.sleep(0.1) with the monkey patch) and in the data-read thread but the kube pod still misses answering liveness and readyness probes enough times to get killed by kubernetes.

I tried using 2 (and more) workers. That solved my probe issue since a worker was always available for them. But every worker also now launches the read event, and I have multiple gevents trying to do the same thing.

Can I "pin" a gevent to a worker? Like only worker #0 does this event? Should I use gthreads/thread pools instead? I'm not terribly familiar with gunicorn so I'm sorry if this is a silly question.

In C++ I'd just launch a thread to do the cpu intensive task and let the scheduler handle the thread preemption, but I'm not sure what I should do here