Closed jsrikish closed 10 months ago
@jsrikish, @krisstanton, and I went over the Cumulus release notes from the version immediately following 13.4.0 (our current version) through 16.1.1. Offhand, it appears that our upgrade path should be relatively smooth. Here are the pertinent points we extracted from the various release notes:
main
branch and create a new branch named cumulus-upgrade-16.1.1
engine_version
in app/stacks/rds-cluster/main.tf
to match..nvmrc
file)module "cumulus"
block within app/stacks/cumulus/main.tf
:
data "aws_ecr_repository" "async_operation" {
name = "async_operations"
}
module "cumulus"
block within app/stacks/cumulus/main.tf
(note that v15.0.0 sets the version number to 44, but the release notes for v16.1.1 bump it to 47):
async_operation_image = "${data.aws_ecr_repository.async_operation.repository_url}:47"
async_operation_image
is subsumed by change noted in v16.1.1)module "cumulus"
block within app/stacks/cumulus/main.tf
orca_api_uri = module.orca.orca_api_deployment_invoke_url
orca_lambda_copy_to_archive_arn = module.orca.orca_lambda_copy_to_archive_arn
orca_sfn_recovery_workflow_arn = module.orca.orca_sfn_recovery_workflow_arn
app/stacks/cumulus/main.tf
, find module.orca.orca_lambda_copy_to_glacier_arn
and change the name to module.orca.orca_lambda_copy_to_archive_arn
bin/set-cumulus-version.sh 16.1.1
app/stacks/cumulus/orca.tf
, change the version of the orca zip file (value of source
property inside module "orca"
) from v6.0.3
to v8.0.1
(although this ORCA version upgrade is not explicitly stated in the Cumulus v16.1.1 release notes, there is a note that the example file was updated to use ORCA v8.0.1, so we're keeping things aligned with the example)system-bucket
to allow load balancer to post server access logs to the bucket.app/stacks/rds-cluster/main.tf
, add this new block:
resource "random_password" "db_user_password" {
length = 50
upper = true
lower = true
number = true
special = true
}
app/stacks/rds-cluster/main.tf
, change this line:
rds_user_password = random_password.db_password.result
to this:
rds_user_password = random_password.db_user_password.result
(i.e., on the right side of the assignment, change db_password
to db_user_password
, which is the name of the new block added in the previous point)
After making all of the file changes above, do the following in a sandbox deployment:
make all-init
(this might take a fair bit of time because it likely triggers some large zip file downloads)make all-up-yes
(because we made a change in the rds-cluster module in addition to the changes in the cumulus module)