s3-echoer
s3-echoer copied to clipboard
Reads input from stdin and uploads it to an existing S3 bucket
S3 echoer
A simple demo tool that reads input from stdin
and uploads it to an
exiting S3 bucket, keyed by the creation timestamp.
- Prepare S3 bucket
- Use it locally
- Use it in Kubernetes
- Verify S3 write
Prepare S3 bucket
Make sure the bucket you want to write to exists, for example, let's assume we
want to write to a bucket called s3-echoer-demo
. So first we define the target bucket using an environment variable like so:
$ TARGET_BUCKET=s3-echoer-demo
Now you can check if the S3 bucket exists:
$ aws s3 ls | grep $TARGET_BUCKET
If the S3 bucket doesn't exist yet, create it like so:
$ aws s3api create-bucket \
--bucket $TARGET_BUCKET \
--create-bucket-configuration LocationConstraint=$(aws configure get region) \
--region $(aws configure get region)
Locally
Install it
To install s3-echoer
locally, execute the following two commands. Download the
respective binary for your platform from the releases page,
here shown for macOS
, and make it executable:
$ curl -L https://github.com/mhausenblas/s3-echoer/releases/latest/download/s3-echoer-macos -o /usr/local/bin/s3-echoer
$ chmod +x /usr/local/bin/s3-echoer
Use it
To write to the S3 bucket from your local machine, use s3-echoer
as follows:
$ s3-echoer $TARGET_BUCKET
This is a test.
And it should land in the target bucket ...
CTRL+D
Uploading user input to S3 using s3-echoer-demo/s3echoer-1563906471
Kubernetes
Running s3-echoer
in-cluster means that we somehow need to give it the
permission to write to S3. There are two approaches to this: granting the node
the pod runs on the permission or granting the pod the permission, directly.
Node-level approach
With the node-level approach, we grant all nodes the permission to write to S3.
For this, we need to look up the respective role and attach the AmazonS3FullAccess
policy like shown in the following.
The EKS cluster named s3echoer
was provisioned with eksctl
and hence we're looking
for a node group called eksctl-s3echoer-nodegroup-ng
:
$ INSTANCE_PROFILE_PREFIX=$(aws cloudformation describe-stacks | jq -r '.Stacks[].StackName' | grep eksctl-s3echoer-nodegroup-ng)
$ INSTANCE_PROFILE_NAME=$(aws iam list-instance-profiles | jq -r '.InstanceProfiles[].InstanceProfileName' | grep $INSTANCE_PROFILE_PREFIX)
$ ROLE_NAME=$(aws iam get-instance-profile --instance-profile-name $INSTANCE_PROFILE_NAME | jq -r '.InstanceProfile.Roles[] | .RoleName')
$ aws iam attach-role-policy \
--role-name $ROLE_NAME \
--policy arn:aws:iam::aws:policy/AmazonS3FullAccess
Next, create the service account s3-echoer
that the app will be using:
$ kubectl create sa s3-echoer
Now you can launch the job like so:
$ sed -e "s/TARGET_BUCKET/${TARGET_BUCKET}/g" s3-echoer-job.yaml.template > s3-echoer-job.yaml
$ kubectl apply -f s3-echoer-job.yaml
You can check how the job did using:
$ kubectl logs job/s3-echoer
Uploading user input to S3 using s3-echoer-demo/s3echoer-1565024447
Obviously, this approach has a critical security drawback: it grants every pod running on any of the nodes the right to write to S3, violating the least-privileges principle.
We can do better with the pod-level approach.
Pod-level approach
Using IAM roles for service accounts (IRSA) as per #23 we can grant individual apps certain permissions. Let's do that for our case.
Again, assuming the EKS cluster named s3echoer
was provisioned with eksctl
in version 0.5.0 or above, we first enable IRSA in the cluster like so:
$ eksctl utils associate-iam-oidc-provider --name s3echoer --approve
And now we can create the service account s3-echoer
used by the app and along
with it create a role that has the AmazonS3FullAccess
policy attached:
$ eksctl create iamserviceaccount \
--name s3-echoer \
--cluster s3echoer \
--attach-policy-arn arn:aws:iam::aws:policy/AmazonS3FullAccess \
--approve
Now we have the permissions in place, the pod needs to acquire the temporary
credentials via STS.
This can be achieved either via using an IRSA-enabled SDK
as is the case here,
or manually, by calling aws sts assume-role-with-web-identity
like so:
$ JQ=/usr/bin/jq && curl https://stedolan.github.io/jq/download/linux64/jq > $JQ && chmod +x $JQ
$ curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py && python get-pip.py && pip install awscli --upgrade
$ aws sts assume-role-with-web-identity \
--role-arn $AWS_ROLE_ARN \
--role-session-name mh9test \
--web-identity-token file://$AWS_WEB_IDENTITY_TOKEN_FILE \
--duration-seconds 1000 > /tmp/irp-cred.txt
$ export AWS_ACCESS_KEY_ID="$(cat /tmp/irp-cred.txt | jq -r ".Credentials.AccessKeyId")"
$ export AWS_SECRET_ACCESS_KEY="$(cat /tmp/irp-cred.txt | jq -r ".Credentials.SecretAccessKey")"
$ export AWS_SESSION_TOKEN="$(cat /tmp/irp-cred.txt | jq -r ".Credentials.SessionToken")"
$ rm /tmp/irp-cred.txt
Now you can launch the job like so:
$ sed -e "s/TARGET_BUCKET/${TARGET_BUCKET}/g" s3-echoer-job.yaml.template > s3-echoer-job.yaml
$ kubectl apply -f s3-echoer-job.yaml
You can check how the job did using:
$ kubectl logs job/s3-echoer
Uploading user input to S3 using s3-echoer-demo/s3echoer-1565024447
Note: you can clean up with kubectl delete job/s3-echoer
.
Verify S3 write
Let's check if the data landed in the right place:
$ aws s3api list-objects \
--bucket $TARGET_BUCKET \
--query 'Contents[].{Key: Key, Size: Size}'
[
{
"Key": "s3echoer-1563907403",
"Size": 27
}
]
Yay, that worked like a charm! Seems an S3 object with our content has been created, in the target bucket. And as a final check, let's download the object and store it in a file to check if it actually contains the text the Kubernetes job created:
$ aws s3api get-object \
--bucket $TARGET_BUCKET \
--key s3echoer-1563907403 \
/tmp/s3echoer-readback.txt
$ cat /tmp/s3echoer-readback.txt
This is an in-cluster test
And that's it :)