terraform-aws-notify-slack: Unable to load notify_slack.zip (no such file or directory)

I ran into this issue and found a permanent solution that doesn’t require modifying the module.

The issue is as @stephgosling describes. The module’s path is hard coded into the state. This is a problem because in the TF registry, each version of the module has a unique sha affixed path. So any time the module version changes, it will break. The only way to fix it, is to manually terraform taint the lambda function resource. Again, you must do this every time the module changes.

However, this problem is caused exclusively by using the module via the terraform registry. Using any other module source (e.g. github) does not have this issue because the module path will not differ version to version, just the actual content (which is the only thing that matters)

So to fix this issue permanently:

1. Change module source from terraform registry to github (e.g. source = "github.com/terraform-aws-modules/terraform-aws-notify-slack.git?ref=v2.4.0")
2. Manually taint the resource e.g. terraform taint "<nested-resource-path>.module.notify_slack.aws_lambda_function.notify_slack[0]"
3. Run terraform plan/apply

Now if you change module version, it will update properly without manual intervention

In my opinion updating a module should never lead to a problem with a previously made state, this zipfile is the first i’ve seen that happen.

For folks running separate plan/applies in ephemeral CI/CD and not passing the .terraform directory between the two, this appears to be the cause of the issue.

I just ran into the same issue, fixed by removing this line: https://github.com/terraform-aws-modules/terraform-aws-notify-slack/blob/master/main.tf#L80

FYI line 80 in main.tf is a moving target, I think you were referring to this specific version: https://github.com/terraform-aws-modules/terraform-aws-notify-slack/blob/1d2381abf0e0fc0a0159f06a3aba98b712520661/main.tf#L80

Removing filename from the lifecycle looks like this:

  lifecycle {
    ignore_changes = [
      "last_modified",
    ]
  }

And for what its worth that solved this problem for me as well. However it was added to address: https://github.com/terraform-aws-modules/terraform-aws-notify-slack/issues/6

I see, sometimes when you have to deal with existing (often corrupted) information in state file you can try these workarounds (starting from the easiest at the top):

1. Append -refresh=false to prevent it from updating the state before planning changes.
2. Append -target=module.databases.module.rds_notify_slack.module.notify_slack.aws_lambda_function.notify_slack to limit the scope of changes you want to plan and apply.
3. Use terraform taint within a module.
4. If you don’t mind doing surgery on json file (which is often not needed), you can use terraform state pull, do changes, terraform state push. This is the most advanced way, so I put it at the end.

Let’s see what @cgomestw thinks.