I have a fairly simple micro service that reads data from a MongoDB cluster, does some data transformation, and exposes the data through a REST API. I need to update the independent persistence datastore using a cron job. I could create a separate application to update the dataset but it is easier to deploy just the one application that exists in an AWS auto scaling group (this is for a large enterprise with a lot of red tape for releasing new applications), and have one of the instances update the dataset through a background job.
Locking writes to the DB through a field in the DB is a workable solution but seems like an antipattern. Is there a better way to do this without creating a separate application to do the DB writes?
I think creating or not a new application is a minor question here. It just depends on your architecture. If you develop a microservice, you might want to isolate and remove from it everything that's possible to keep it from making a monolythic again. At least I'd do that. It may seem easier now to keep all in one application, but I will be less so once your codebase and functionality grows considerably.
The question that bothers me even more in this situation is exclusive DB locking. Avoid exclusive DB locks for all costs. Forget about everything, drop all your tasks and find a way to compute your statistics without locking database. Believe me, whatever time you spend now developing a solution will be rewarded in the future.
If you choose to ask this question here on Stack, I would be happy to assist, but to suggest something, we will need to know more about your logic and data you store / aggregate.