grpc: received message larger than max when "upload" references large files

Terraform Version

0.12.19

Affected Resource(s)

docker_container

Terraform Configuration Files

resource "docker_container" "app-core" {
  image = var.tomcat_base_image
  name = "${var.slot}-app-core"

  upload {
    file = "/opt/tomcat/webapps/b.war"
    content_base64 = filebase64(var.first_war_path)
  }
  upload {
    file = "/opt/tomcat/webapps/g.war"
    content_base64 = filebase64(var.second_war_path)
  }
  upload {
    file = "/opt/tomcat/conf/Catalina/localhost/b.xml"
    content_base64 = filebase64(var.tomcat_context_path)
  }
  upload {
    file = "/opt/tomcat/conf/Catalina/localhost/g.xml"
    content_base64 = filebase64(var.tomcat_context_path)
  }

  networks_advanced {
    name = "frontend"
  }
  networks_advanced {
    name = var.slot
    aliases = [
      "<snip>"
    ]
  }

  env = [for k,ec in merge(
    local.app-core-env-config,
    local.shared-env-config
  ): "${k}=${ec}"]
}

Debug Output

There's a lot of info that I'm not keen on sharing, but this is likely the only thing relevant:

2020/01/13 18:51:33 [DEBUG] Resource instance state not found for node "docker_container.app-core", instance docker_container.app-core
2020/01/13 18:51:33 [DEBUG] ReferenceTransformer: "docker_container.app-core" references: []
2020/01/13 18:53:39 [ERROR] <root>: eval: *terraform.EvalDiff, err: rpc error: code = ResourceExhausted desc = grpc: received message larger than max (1023714897 vs. 4194304)
2020/01/13 18:53:39 [ERROR] <root>: eval: *terraform.EvalSequence, err: rpc error: code = ResourceExhausted desc = grpc: received message larger than max (1023714897 vs. 4194304)

Panic Output

Expected Behavior

I assume the war files should have been copied.

Actual Behavior

Error: rpc error: code = ResourceExhausted desc = grpc: received message larger than max (1023714897 vs. 4194304)

Steps to Reproduce

Please list the steps required to reproduce the issue, for example:

terraform apply with a docker_container that has a large file uploaded to it

Important Factoids

Nope

References

Likely related to https://github.com/hashicorp/terraform/issues/21709

As an aside, I would argue that if possible we shouldn't be keeping the entire content_base64 as a state item if possible. A sha512 or some other hash should be more than enough to know whether or not it's changed, and regardless of the content if there is a change, the entire thing is recreated anyway. Just wish I actually know Go enough to be of some help here.

hashicorp / terraform-provider-docker