1 - build lambda locally 2 - deploy using built 'Base2.Lambdas.zip' package (manaully for now) 3 - run the function to generate report. use payload from test/run section of this README to see parameters 4 - run the function to cleanup orphaned AMI EBS snapshosts
You will need docker engine and zip
utility to build project. Also, build script uses bash
shell
If you have dotnet
cli locally installed you may use scripts/build_native.sh
, but docker build is
recommended way for automating builds.
$ scripts/build_docker.sh
Restoring packages for /project/Base2.Lambdas.csproj...
Lock file has not changed. Skipping lock file write. Path: /project/obj/project.assets.json
Restore completed in 2.06 sec for /project/Base2.Lambdas.csproj.
NuGet Config files used:
/root/.nuget/NuGet/NuGet.Config
Feeds used:
https://api.nuget.org/v3/index.json
Microsoft (R) Build Engine version 15.1.1012.6693
Copyright (C) Microsoft Corporation. All rights reserved.
Base2.Lambdas -> /project/bin/Debug/netcoreapp2.1/Base2.Lambdas.dll
adding: AWSSDK.AutoScaling.dll (deflated 70%)
adding: AWSSDK.Core.dll (deflated 66%)
adding: AWSSDK.EC2.dll (deflated 70%)
adding: AWSSDK.S3.dll (deflated 63%)
adding: Amazon.Lambda.Core.dll (deflated 57%)
adding: Amazon.Lambda.Serialization.Json.dll (deflated 56%)
adding: Base2.Lambdas.deps.json (deflated 74%)
adding: Base2.Lambdas.dll (deflated 55%)
adding: Base2.Lambdas.pdb (deflated 40%)
adding: Newtonsoft.Json.dll (deflated 60%)
adding: System.Collections.NonGeneric.dll (deflated 60%)
adding: System.Runtime.Serialization.Primitives.dll (deflated 48%)
You will need serverless framework, version > 1.15
to deploy lambda functions automatically. Use sls deploy
,
in comnbination with properly set environment variables:
$ export REGION=ap-southeast-2
$ export SOURCE_BUCKET=automation.cleanup.base2.services
$ sls deploy
Serverless: Packaging service...
Serverless: Uploading CloudFormation file to S3...
Serverless: Uploading artifacts...
Serverless: Validating template...
Serverless: Creating Stack...
Serverless: Checking Stack create progress...
.........................................
Serverless: Stack create finished...
Service Information
service: manualawscleanup
stage: dev
region: ap-southeast-2
api keys:
None
endpoints:
None
functions:
AMIReport: manualawscleanup-dev-AMIReport
AMICleanup: manualawscleanup-dev-AMICleanup
EBSReport: manualawscleanup-dev-EBSReport
EBSCleanup: manualawscleanup-dev-EBSCleanup
Note that all of configurtion below is now implemented through serverless framework, and thus
scripts/build_docker.sh
script will create lambda package in root directory called Base2.Lambdas.zip
.
This package is referenced in serverless project as code package.
Use following entry points (Lambda function handlers)
- Report generation for EBS -
Base2.Lambdas::Base2.Lambdas.Handlers.EBSReportAndCleanup::UploadEBSReport
- Report generation for AMI -
Base2.Lambdas::Base2.Lambdas.Handlers.AMIReportAndCleanup::UploadAMIReport
- Cleanup from CSV info for EBS -
Base2.Lambdas::Base2.Lambdas.Handlers.EBSReportAndCleanup::CleanupFromReport
- Celanup from CSV info for AMIs -
Base2.Lambdas::Base2.Lambdas.Handlers.AMIReportAndCleanup::DeregisterReportedAMIs
Iam role configured for lambda should have following policies
- read only access to EC2 service
- write acces to S3 bucket passed in as argument
{
"Version": "2012-10-17",
"Statement": [
{
"Sid": "Stmt1497509441000",
"Effect": "Allow",
"Action": [
"s3:*"
],
"Resource": [
"arn:aws:s3:::aws.amis-cleanup.reports.example.com/*"
]
}
]
}
- DeleteSnapshot permissions
{
"Version": "2012-10-17",
"Statement": [
{
"Sid": "Stmt1497854974000",
"Effect": "Allow",
"Action": [
"ec2:DeleteSnapshot"
],
"Resource": [
"*"
]
}
]
}
- Invoke lambda permission, to invoke itself recursively for long running deletions
{
"Version": "2012-10-17",
"Statement": [
{
"Sid": "Stmt1497921290000",
"Effect": "Allow",
"Action": [
"lambda:InvokeAsync",
"lambda:InvokeFunction"
],
"Resource": [
"*"
]
}
]
}
All of operations can be time consuming, so it's recommended to set all runtimes to 5 minutes
Use C# as runtime
This functions do not require more than 128MB of memory, even when working with ~10k EBS snapshots (highest tested value)
There is no VPC configuration required
Both report generation and cleanup tasks are accepting location of csv file to write/read
in event parameters. For report generation there is optional parameter OnlyAMIOrphans
which default to
true
. This parameter determines whether only AMI orphans get reported or ALL EBS snapshots
(danger zone, as you don't want to delete all snapshots, but you may want to delete some that are not
orphans, thus need for this functionality)
e.g.
{
"BucketName":"aws.amis-cleanup.reports.base2.services",
"Key":"ebs_report_prod.csv",
"OnlyAMIOrphans": true
}