https://www.growthbook.io/
Join Slack
I am setting up GrowthBook <Proxy> and the documentation does not seem clear on various points. It m...
# ask-questions
c
I am setting up GrowthBook Proxy and the documentation does not seem clear on various points. It mentions 
SECRET_API_KEY
and to setup in Settings -> API Keys. Does that need to be an 
Admin
key or can it be just a 
Read-Only
key?
h
A read-only key is fine. I can update the docs to make this more clear.
c
thanks
If you do update the docs (which would be great) it would nice nice to know which variables are options and which are required. Also if there are default values that would be helpful. I assume cache will default to memory if not set.
h
I'll have a look. Regarding cache engine, that is correct: the default is memory. In the near term, you can check out 
packages/apps/proxy/src/init.ts
to see what the defaults are.
👍🏼 1
c
Any idea what might cause this error in the proxy log?
Copy code
| {"level":50,"time":1697047211013,"pid":67,"hostname":"ip-10-163-18-220.eu-west-1.compute.internal","err":{"type":"TypeError","message":"fetch failed: Connect Timeout Error","stack":"TypeError: fetch failed\n at Object.fetch (node:internal/deps/undici/undici:11576:11)\n at process.processTicksAndRejections (node:internal/process/task_queues:95:5)\ncaused by: ConnectTimeoutError: Connect Timeout Error\n at onConnectTimeout (node:internal/deps/undici/undici:8522:28)\n at node:internal/deps/undici/undici:8480:50\n at Immediate._onImmediate (node:internal/deps/undici/undici:8511:13)\n at process.processImmediate (node:internal/timers:476:21)"},"msg":"polling error"}
I verified the proxy can reach the API server via telnet.
packet capture from proxy looks like things are returning 200 OK.
h
link isn't opening, but I don't have a good sense of why you'd get an error on this endpoint (yet the GB API has returned a 200). I'm assuming you've set 
SECRET_API_KEY
appropriately?
are you running the dockerized app? What server environment?
c
oops did not mean to post the DNS name. It is not publically routable
👍 1
Yes it is dockerized app. This is part of the CDK to bring it up.
Copy code
const proxyContainerName = createResourceName('dna-growthbook-proxy', RESOURCE_TYPE.EC2_INSTANCE, this.app.project)

    growthbookProxyTask.addContainer('DnaGrowthBookProxyContainer', {
      image: ecs.ContainerImage.fromRegistry('growthbook/proxy:latest'),
      containerName: proxyContainerName,
      essential: true,
      logging: ecs.LogDrivers.awsLogs({
        streamPrefix: 'blah-growthbook-proxy',
        logGroup,
      }),
      environment: proxyEnnvironment,
      secrets: proxySecrets,
      portMappings: [{
        containerPort: 3300,
      }],
    })
It all seems to be working.... I just don't know why we see some error in the logs.
h
so the proxy works? you just randomly see this error sometimes?
c
Yes
Also seeing this:
Copy code
2023-10-11T10:16:34.242-05:00	yarn run v1.22.19	Link 
2023-10-11T10:16:34.338-05:00	$ wsrun -p @growthbook/proxy -c start	Link 
2023-10-11T10:16:35.519-05:00	@growthbook/proxy	Link 
2023-10-11T10:16:35.519-05:00	| $ node dist/index.js	Link 
2023-10-11T10:16:36.627-05:00	| (node:70) [DEP0111] DeprecationWarning: Access to process.binding('http_parser') is deprecated.	Link 
2023-10-11T10:16:36.627-05:00	| (Use `node --trace-deprecation ...` to show where the warning was created)	Link 
2023-10-11T10:16:36.629-05:00	| GrowthBook proxy running over HTTP1.1, port 3300	Link 
2023-10-11T10:16:36.644-05:00	| /usr/local/src/app/packages/apps/proxy/dist/services/registrar/index.js:152	Link 
2023-10-11T10:16:36.702-05:00	| throw new Error("missing required context for polling for connections");	Link 
2023-10-11T10:16:36.702-05:00	| ^	Link 
2023-10-11T10:16:36.702-05:00	|	Link 
2023-10-11T10:16:36.703-05:00	| Error: missing required context for polling for connections	Link 
2023-10-11T10:16:36.703-05:00	| at /usr/local/src/app/packages/apps/proxy/dist/services/registrar/index.js:152:19	Link 
2023-10-11T10:16:36.703-05:00	| at Generator.next (<anonymous>)	Link 
2023-10-11T10:16:36.703-05:00	| at /usr/local/src/app/packages/apps/proxy/dist/services/registrar/index.js:8:71	Link 
2023-10-11T10:16:36.703-05:00	| at new Promise (<anonymous>)	Link 
2023-10-11T10:16:36.704-05:00	| at __awaiter (/usr/local/src/app/packages/apps/proxy/dist/services/registrar/index.js:4:12)	Link 
2023-10-11T10:16:36.705-05:00	| at initializeRegistrar (/usr/local/src/app/packages/apps/proxy/dist/services/registrar/index.js:130:42)	Link 
2023-10-11T10:16:36.705-05:00	| at /usr/local/src/app/packages/apps/proxy/dist/app.js:99:47	Link 
2023-10-11T10:16:36.705-05:00	| at Generator.next (<anonymous>)	Link 
2023-10-11T10:16:36.706-05:00	| at /usr/local/src/app/packages/apps/proxy/dist/app.js:31:71	Link 
2023-10-11T10:16:36.707-05:00	| at new Promise (<anonymous>)	Link 
2023-10-11T10:16:36.707-05:00	|	Link 
2023-10-11T10:16:36.707-05:00	| Node.js v18.18.0	Link 
2023-10-11T10:16:36.748-05:00	| error Command failed with exit code 1.	Link 
2023-10-11T10:16:36.748-05:00	| info Visit <https://yarnpkg.com/en/docs/cli/run> for documentation about this command.	Link 
2023-10-11T10:16:36.779-05:00	| `yarn start` failed with exit code 1	Link 
2023-10-11T10:16:36.818-05:00	error Command failed with exit code 1.	Link
h
that error indicates that your proxy instance does not have 
GROWTHBOOK_API_HOST
and 
SECRET_API_KEY
env vars
which could also be consistent with the other polling error you're seeing above
c
It does have those set. And the proxy is working...
Maybe this was just an issue when I first brought it up. Will play with it some more and see if it keeps happening.
Thanks for the tips
h
So the proxy will "work" (keep running, minimally responding to input) if these fields aren't set. But it won't do anything useful like caching or streaming.
One thought: I did just ship a large refactor of the codebase a few days ago. If you want to pull the previous tagged release before this refactor, we could see if that helps https://github.com/growthbook/growthbook-proxy/releases/tag/v1.0.26
(of course that does complicate things a bit, harder to spin up)
c
We are testing auto scaling at the moment. But will try and pull older version later today and test it out
When I do 
env
in the shell the env variables are for sure set...
Copy code
> env
AWS_EXECUTION_ENV=AWS_ECS_FARGATE
AWS_CONTAINER_CREDENTIALS_RELATIVE_URI=/v2/credentials/...
HOSTNAME=ip-10-163-9-nnn.compute.internal
YARN_VERSION=1.22.19
PWD=/usr/local/src/app
ECS_CONTAINER_METADATA_URI_V4=<http://169.254.170.2/v4/xxx>
HOME=/root
LANG=C.UTF-8
GROWTHBOOK_API_HOST=<http://api.growthbook.local:3100>
CACHE_ENGINE=memory
ECS_AGENT_URI=<http://169.254.170.2/api/xxx>
TERM=xterm-256color
ENVIRONMENT=playground
ECS_CONTAINER_METADATA_URI=<http://169.254.170.2/v3/xxx>
SHLVL=1
SECRET_API_KEY=secret_readonly_xxx
PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
NODE_VERSION=18.18.0
_=/usr/bin/env
Just replaced some potentially sensitive info with 
xxx
or 
...
Is there any recommendations for when to scale up to more instances of proxies. Is it usually CPU, Memory, connections or something else that tend to exhaust first on a given instance?
h
internally we base it on CPU 
ECSServiceAverageCPUUtilization
at 80
c
perfect
That is what we just tested and seemed to work well.
I think we will add this too just in case:
Copy code
scalableTaskCount.scaleOnCpuUtilization('CpuUtilizationScaling', {
    targetUtilizationPercent: 80,
})
scalableTaskCount.scaleOnMemoryUtilization('MemoryScaling', {
    target_utilization_percent=80,
})
90 Views
Open in Slack
PreviousNext
Powered by

GrowthBook Users

Open source platform for stress free deployments, measured impact, and smarter decisions.

Powered by