I’ve been having issues w/ unexpected 502 errors all day on my node server. Although resources are nowhere near limits, autoscaling from 1 (~40 req/instance) to 10 instances (~4 req/instance) helped my service work faster. Still, we’re getting a lot of 502 errors:
Also, these 502 err codes are not showing in Grafana:
Grafana appears to be showing the “app” responses count, while in our UI we show the “edge” responses count. That includes errors from our proxy. That’s an inconsistency and a mistake on our end.
Looking at the 502s we’re returning, there are a few things I noticed:
- Your app becomes unhealthy frequently. Perhaps tweaking the health checks would help
- We often get connection timeouts connecting to your app
- Response time is high
- System metrics for your instances look fine
This leads me to believe you might have to make sure your server is efficient at accepting connections and isn’t blocking for whatever reason.