From @xmunoz's comments in slack:
Running database migrations during image building is causing intermittent failure. I believe this is because the images aren't guaranteed to be built on the same subnets are the running instances, which means that they sometimes won't have network access to the DBs. It might be possible to define the subnet images get built on, but in the mean time, I'm going to remove db migrations from the deployment process. (Done in 535eb99)
Going forward, we would like to add migrations back into our deployment script.
subnet_id or vpc_id might be the solution. Add one of these to the image config: https://www.packer.io/docs/builders/amazon/ebs#vpc_id
This will require specifying the value explicitly in the server image config, and passing that variable through to image.json: https://github.com/PermanentOrg/infrastructure/tree/main/images
From @xmunoz's comments in slack:
Running database migrations during image building is causing intermittent failure. I believe this is because the images aren't guaranteed to be built on the same subnets are the running instances, which means that they sometimes won't have network access to the DBs. It might be possible to define the subnet images get built on, but in the mean time, I'm going to remove db migrations from the deployment process. (Done in 535eb99)
Going forward, we would like to add migrations back into our deployment script.
subnet_id or vpc_id might be the solution. Add one of these to the image config: https://www.packer.io/docs/builders/amazon/ebs#vpc_id
This will require specifying the value explicitly in the server image config, and passing that variable through to image.json: https://github.com/PermanentOrg/infrastructure/tree/main/images