Without this, and considering we had only a single redis instance backing both the production and development deployments, it was arbitrary which deployment would pick up enqueued tasks (dev vs production). So a worker from the dev deployment would randomly pick up a task enqueued by the production spackbot webservice, or vice versa, and in this case it would not have any of the correct secrets to successfully complete the job. This change makes the queue name an environment variable so the two deployment environments can use distinct work queues in the same redis instance.
scottwittenburg
commented
2 years ago
Without this, and considering we had only a single redis instance backing both the production and development deployments, it was arbitrary which deployment would pick up enqueued tasks (dev vs production). So a worker from the dev deployment would randomly pick up a task enqueued by the production spackbot webservice, or vice versa, and in this case it would not have any of the correct secrets to successfully complete the job. This change makes the queue name an environment variable so the two deployment environments can use distinct work queues in the same redis instance.