has anyone experienced an issue with a fully manually created ECS Cluster where the instance during it's creation does not have network access, but does later on ? I went through every bit of my config:
• Security Group allow eggress anywhere
• the instance is in the private subnet, which routing table for outbound traffic points to the NAT
• The NAT Gateway is in public subnet
• the pub subnet does have an IG
Additionally I created a jumbox and while the stack fails to create (no network so no cfn-signal installed by cloud-init and sent to Cloud formation for the autoscaling group). Once the instance is up, and without changing anything I log in to that specific and can successfully reach internet... confirming that the setup is right.
I am thinking that the instance comes up before the NAT is attached or a specific rule added to routing table or Security group. Any pointer would be welcomed
06/14/2022, 12:58 PM
there's lots of eventual consistencies like this in AWS 😅
You can do things like run a loop inside the user data
ping -c 1 ip
Or the pulumi specific way is to use a waiter on certain resources
06/14/2022, 1:01 PM
never had the issue with cloudformation, but I guess they are aware of their own things lool.
now that I know how I can specify my own userData (and have the instance reach out to cfn & add itself to the ECS cluster) I can do something like that yes, and I may actually use the dependsOn to make sure things get created in a certain order