Hi everyone, I have a few questions: - How can we ...
# ask-questions
a
Hi everyone, I have a few questions: • How can we check the number of API requests that we have used out of the limit? • I have fully rolled out a feature on FE (tried 100% rollout rule and a default value without any other rules), but some of the requests (5-10%) are still going to the old endpoint - what are some common causes that lead to this behaviour? Will setting up our own GrowthBook Proxy help? Thanks!
w
• How can we check the number of API requests that we have used out of the limit?
Unfortunately we don't have a way for you to check that at the moment. If you DM me the email of the owner of the account I can look it up for you.
• I have fully rolled out a feature on FE (tried 100% rollout rule and a default value without any other rules), but some of the requests (5-10%) are still going to the old endpoint - what are some common causes that lead to this behaviour? Will setting up our own GrowthBook Proxy help?
Changing default cache settings, or clients on mobile or some other environment where streaming updates might not be working correctly without setting the right polyfills are two things that come to mind.
Will setting up our own GrowthBook Proxy help?
The proxy can control how many api requests you are using, but it shouldn't be changing what eventually gets served in any way.
a
Changing default cache settings, or clients on mobile or some other environment where streaming updates might not be working correctly without setting the right polyfills are two things that come to mind.
The setup I’m using is a ReactJS v17 web app with a myriad of users on different devices. In this case, will polyfills be required? Another piece of information that might be helpful is that we’re proxying the requests through a Cloudflare worker script.
w
The setup I’m using is a ReactJS v17 web app with a myriad of users on different devices. In this case, will polyfills be required?
No.
Another piece of information that might be helpful is that we’re proxying the requests through a Cloudflare worker script.
You are proxying the Growthbook features endpoint through Cloudflare?
1
Cloudflare could indeed not be clearing the cache appropriately. You can also use Growthbook Proxy: https://docs.growthbook.io/self-host/proxy.
👀 1
Are you seeing these 5%-10% consistently, or are they disappearing after awhile (like some cache expiring)
a
Rather consistently, but since the caches have 1 min TTL, and we have many clients hitting growthbook, I’m not sure if I can see whether it’s because of the cache expiring?
w
If you are still seeing 5-10% of your users hitting the old value for a feature that is 100% rolled out after the 1 min TTL, then perhaps your cloudFlare or something is getting in the way. You can use the Growthbook DevTools extension to help debug: https://chrome.google.com/webstore/detail/growthbook-devtools/opemhndcehfgipokneipaafbglcecjia.
a
Oh it works perfectly on local though, and the DevTools is used to debug on local, right? Or do you mean I should be load testing my local / staging to see if 5-10% of it fails? 😅
w
You can use the devtools debugger to analyze the results from the features endpoint to make sure it is what you expect. It works on prod.
a
Seems like Dev mode needs to be on? Or am I missing something?
w
ah right. Sorry. I led you astray again. 😓
Well you can look at the features endpoint in normal dev console and should be able to verify the result that way.
My guess is that you will see rules still that would allow some people to see it 5-10% of the time. So then the question is whether CloudFlare has an old cache, or perhaps your Growthbook feature that you thought you were configuring for prod is actually still the configuration for dev. (Some people have been confused by the UI there)
Some people forgot to click the right tab here:
a
No, we have configured the right environment. Also, we have changed the default to true and removed all rules - the same is observed. I suspect that growthbook sdk is not loading the feature definitions for some reason, possibly due to network errors or their VPN settings.
Here are our status codes for the past 24h - what can possibly be causing the non-200 requests? 😅
w
Those are the status codes of hitting cloudFlare url, or of cloudFlare hitting cdn.growthbook.io?