This message was deleted.

Sounds like a network issue with the EKS cluster and the worker nodes. Have you double checked your security groups to verify egress rules?

Have you tried the same operation without Atlantis? Running plan or apply locally?

Yes egress has route for internet connectivity, also I am calling my module itself from another Bitbucket repo which is working just fine

locally from the pod ?

Try the usual method of breaking the problem down. If it works locally from your machine, but not from the pod, it may be a networking error with your EKS cluster, so check that in isolation. Fire up an EKS pod that has telnet available (or whatever utility you like using) and check if you can establish a connection, in a vacuum from Helm, Terraform, and Atlantis —

telnet 185.199.109.153 443

.

@Gabor Maghera but my other modules are also pulled from BitBucket/Github but I am facing this issue specifically with Helm provider based resources, if its network issue, other modules sources should fail as well with timeout error

you need to first isolate the issue to Atlantis or terraform.

Try local first, from laptop, and then you can also try emulating what Atlantis does, by exec’ing into the pod, and firing off a plan from the MR directory. Based on what you describe, it may be the provider.

So I was able to fix this, issue was with using custom CNI [cilium] within my cluster and as Atlantis statefulset pod runs with dnsPolicy: ClusterFirst, it for some reason wasn’t allowing http/https curl requests from within pod, switching back to VPC CNI solved the issue, I am now looking at how same behaviour can be achieved with Cilium CNI