This message was deleted.

@blue-hair-70812 has opened an issue Close Issue button

Hi Travis, I’ll sync with our engineers who work on the proxy and get back to you!

thanks!

Hi Travis, I forgot to ask-- Are you using the cloud-hosted or self-hosted version? Which SDK are you using?

self-hosted. sdk is javascript

Hi Travis, I'm troubleshooting this with one of our engineers. Are you getting the error only when clicking "re-check", or when trying to actually run the proxy If you could send some more details about what this is blocking, and maybe a screenshot or two, that would help us a lot!

i am getting the error when "rechecking":

Ahh OK! If the proxy's ​`/healthcheck`​ endpoint returns a ​`proxyVersion: undefined`​ this can happen. And it would be undefined perhaps if using a configuration where the ​`package.json`​ is unavailable for parsing; possibly a docker-specific bug. But it should not affect anything; it doesn't mean anything is actually broken other than the healthcheck itself (which is diagnostic only).

quick question before i try and gather that - for the eventstreaming - does this happen through a websocket?

Let me confirm, just a moment

FYI - we are running inside Kubernetes and i just want to verify that our ingress doesn't need some special configuration outside of the normal HTTP proxying - This error could just be from a misconfig on my end 😥

Ahh OK, no worries! Event streaming happens via Server-Sent Events (SSE) and not websockets. Here's the relevant part from the JavaScript SDK docs: https://docs.growthbook.io/lib/js#streaming-updates

It's*

Hi Travis, is there anything else we can help with on this topic?

Hi @brief-honey-45610 let me bring some light on this issue with SSE When we used cloud integration everything was ok with SSE connection (1 screen) When we moved to self-hosting with proxy solution we lost event stream connection (2 screen) JS SDK integration details: apiHost:“https://growthbook-proxy...” backgroundSync:true clientKey:“...” enableDevMode:true subscribeToChanges: true Do you have any thoughts on which direction we should proceed investigation?

Hi Oleksandr, I apologize for the delay. I'm working with one of our engineers to troubleshoot this today.

Hi @brief-honey-45610 “Did you change the API host that your SDK client connects to?” - Yes to proxy Other questions I will address to our devOps

hey there. We conducted additional tests by adding these headers to GrowthBookFeatures and began to observe preflight requests (wonder if it correlates somehow). Maybe we need to add something else?:

apiHostRequestHeaders: {
        'X-Accel-Buffering': 'no',
        'Content-Type': 'text/event-stream'
      },
      streamingHostRequestHeaders: {
        'X-Accel-Buffering': 'no',
        'Content-Type': 'text/event-stream'
      }

The nginx Ingress configuration is pretty simple:

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  annotations:
    cert-manager.io/cluster-issuer: letsencrypt-prod-cloudflare
    kubernetes.io/ingress.class: nginx
  labels:
    argocd.argoproj.io/instance: growthbook
  name: growthbook-proxy
  namespace: growthbook

spec:
  rules:
  - host: xxxxxxxx
    http:
      paths:
      - backend:
          service:
            name: growthbook-proxy
            port:
              number: 3300
        path: /
        pathType: ImplementationSpecific
  tls:
  - hosts:
    - xxxxxxxx
    secretName: growthbook-proxy-tls

It’s difficult to troubleshoot because neither Nginx, GrowthBook, or the GrowthBook proxy records any errors in logs. 1st screenshot - without headers 2nd - with last one - working one. without nginx

@brief-honey-45610 you can close this issue nginx.ingress.kubernetes.io/proxy-buffering: “off” helped us