From 45af487b2102de89118624655e7f5d4a9da8ac67 Mon Sep 17 00:00:00 2001 From: Stefan Freitag Date: Wed, 18 Oct 2023 20:56:48 +0200 Subject: [PATCH] feat: add support for suppressing notifications (#3) Users can specify MSK cluster states that should not trigger a notification. - Updated pre-commit hooks to use newer versions: - antonbabenko/pre-commit-terraform@v1.83.5 - pre-commit/pre-commit-hooks@v4.5.0 --- .gitignore | 3 ++ .pre-commit-config.yaml | 4 +- README.md | 2 + data.tf | 2 +- examples/01_default_configuration/README.md | 2 +- functions/check-msk-status/index.py | 56 ++++++++++++--------- main.tf | 20 ++++---- variables.tf | 21 ++++++++ 8 files changed, 73 insertions(+), 37 deletions(-) diff --git a/.gitignore b/.gitignore index f63ef01..04c1c20 100644 --- a/.gitignore +++ b/.gitignore @@ -26,3 +26,6 @@ terraform.rc # Python virtual environment .venv + +# Lambda zip directory +out/ diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml index db118f6..91861ed 100644 --- a/.pre-commit-config.yaml +++ b/.pre-commit-config.yaml @@ -1,6 +1,6 @@ repos: - repo: https://github.com/antonbabenko/pre-commit-terraform - rev: v1.83.2 + rev: v1.83.5 hooks: - id: terraform_fmt - id: terraform_validate @@ -27,7 +27,7 @@ repos: - --args=--quiet - --args=--skip-check CKV_AWS_116,CKV_AWS_117,CKV_AWS_173,CKV_AWS_272 - repo: https://github.com/pre-commit/pre-commit-hooks - rev: v4.4.0 + rev: v4.5.0 hooks: - id: check-merge-conflict - id: end-of-file-fixer diff --git a/README.md b/README.md index 9281459..0518178 100644 --- a/README.md +++ b/README.md @@ -51,6 +51,8 @@ No modules. | Name | Description | Type | Default | Required | |------|-------------|------|---------|:--------:| | [email](#input\_email) | List of e-mail addresses subscribing to the SNS topic. Default is empty list. | `list(string)` | `[]` | no | +| [ignore\_states](#input\_ignore\_states) | Suppress warnings for the listed MSK states. Default: ['MAINTENANCE'] | `list(string)` | 
[
  "MAINTENANCE"
]
| no | +| [log\_retion\_period\_in\_days](#input\_log\_retion\_period\_in\_days) | Number of days logs will be retained. Default is 365 days. | `number` | `365` | no | | [schedule\_expression](#input\_schedule\_expression) | The schedule expression for the CloudWatch event rule. Default is 'rate(15 minutes)'. | `string` | `"rate(15 minutes)"` | no | | [tags](#input\_tags) | A map of tags to add to all resources. Default is empty map. | `map(string)` | `{}` | no | diff --git a/data.tf b/data.tf index 361d68d..ead1168 100644 --- a/data.tf +++ b/data.tf @@ -8,5 +8,5 @@ data "aws_region" "current" {} data "archive_file" "status_checker_code" { type = "zip" source_dir = "${path.module}/functions/check-msk-status/" - output_path = "${path.module}/python/hello-python.zip" + output_path = "${path.module}/out/check-msk-status.zip" } diff --git a/examples/01_default_configuration/README.md b/examples/01_default_configuration/README.md index 8ceb592..315973f 100644 --- a/examples/01_default_configuration/README.md +++ b/examples/01_default_configuration/README.md @@ -27,4 +27,4 @@ No inputs. ## Outputs No outputs. - \ No newline at end of file + diff --git a/functions/check-msk-status/index.py b/functions/check-msk-status/index.py index 0e5521d..3bb6404 100644 --- a/functions/check-msk-status/index.py +++ b/functions/check-msk-status/index.py @@ -1,37 +1,47 @@ import boto3 import os + def lambda_handler(event, context): - LAMBDASNSTOPIC = os.environ['SNS_TOPIC_ARN'] - region = 'eu-central-1' + LAMBDASNSTOPIC = os.environ["SNS_TOPIC_ARN"] + SUPPRESS_STATES = os.environ["SUPPRESS_STATES"].split(",") + region = "eu-central-1" # Create an MSK client - client = boto3.client('kafka', region_name=region) + client = boto3.client("kafka", region_name=region) # Retrieve a list of clusters response = client.list_clusters() # Extract the cluster ARNs from the response - cluster_arns = response['ClusterInfoList'] + cluster_arns = response["ClusterInfoList"] + + valid_states = ["ACTIVE"] + SUPPRESS_STATES + print( + "Notifications suppressed for these MSK states: {}".format( + ", ".join(valid_states) + ) + ) for cluster in cluster_arns: - arn = cluster['ClusterArn'] + arn = cluster["ClusterArn"] response = client.describe_cluster(ClusterArn=arn) - status = response['ClusterInfo']['State'] - sns_client = boto3.client('sns') - - if status != 'ACTIVE': - print("The MSK cluster: {} needs attention.".format(arn)) - sns_client.publish(TopicArn=LAMBDASNSTOPIC, - Message="MSK cluster: " + arn + " needs attention. The status is: " + status, - Subject="MSK Health Warning!") + status = response["ClusterInfo"]["State"] + print("The cluster is in state {}.".format(status)) + sns_client = boto3.client("sns") + if status not in valid_states: + print("The MSK cluster: {} needs attention.".format(arn)) + sns_client.publish( + TopicArn=LAMBDASNSTOPIC, + Message="MSK cluster: " + + arn + + " needs attention. The status is: " + + status, + Subject="MSK Health Warning!", + ) else: - print( - "The MSK cluster: {} is in a healthy state, and is reachable and available for use.".format( - arn)) + print( + "The MSK cluster: {} is in a healthy state, and is reachable and available for use.".format( + arn + ) + ) # Return the status - return { - 'statusCode': 200, - 'body': 'OK' - } - -if __name__ == '__main__': - lambda_handler(None, None) + return {"statusCode": 200, "body": "OK"} diff --git a/main.tf b/main.tf index 77e477d..98ab7b7 100644 --- a/main.tf +++ b/main.tf @@ -20,8 +20,7 @@ resource "aws_sns_topic_subscription" "msk_health_sns_topic_email_target" { # IAM role resource "aws_iam_role" "msk_health_lambda_role" { - name = "msk-health-lambda-role-${random_id.id.hex}" - + name = "msk-health-lambda-role-${random_id.id.hex}" assume_role_policy = <