Meteor application works very slowly after 1-2 weeks of operation.
The app is deployed in a pod and uses ~30%-50% of the allocated resources. There are not a lot of users,max 10 users simultaneously and ~50 per day.
The Meteor backend is used mainly as a BFF calling Meteor methods (and I don’t do unblock because use Meteor.async), there are still only two subscriptions nd one of them is embedded users subscription for the Users collection.
The chains of requests/responses from client → BFF → API → BFF works quickly and only the last step: from BFF → client takes a lot time.
The issue is gone away if restart the pod and returns again after 1-2 weeks.
I use the pod with NGINX as proxy server in front of the pod with the meteor app.We see a lot of socket connection in TIME_WAIT state on the pod with the meteor app. I saw a lot of messages about the issue with TIME_WAIT sockets and there were write years ago. Is there solution for it?
I have three meteor applications with similar symptoms.
I suspect there are some issue with deploying (custom build pipeline, not MUP) or kubernetes/pods tuning (Ingress, Nginx, something else).
Have you tried to use MontiAPM to see were the slow points are?
Not yet. The security policy does not allow use external service (. I’ll try to use meteor-elastic-apm.