Deployment DNS switchover behavior
We're running into issues setting up continuous deployment with our service. We have a webservice that has long running requests (take up to a couple minutes). We set up RAILWAY_DEPLOYMENT_OVERLAP_SECONDS to make sure the request finishes processing before the old server is removed.
The behavior we're noticing, however, is that during the OVERLAP period, requests are still being routed to the old deployment, rather than beginning to be sent to the new server. Then when the server gets taken down, it cancels all the requests that are ongoing. Is this expected behavior?
Our expectation is that when the RAILWAY_DEPLOYMENT_OVERLAP_SECONDS period begins, Railway would start routing new requests to the new deployment, and let the old deployment finish any outstanding requests in the system, but this seems to not be the case.
4 Replies
Project ID:
95c304a2-fe2b-4ad1-bc3f-5296fd26f36c
95c304a2-fe2b-4ad1-bc3f-5296fd26f36c
Brody! My man
reading now