The good news: AWS announced DynamoDB backups at re:Invent 2017. This will save a lot of unnecessary ops burden.
The bad news? You can't schedule and automate your backups. You need to manually click in the console to create your backup.
Have no fear, an automated solution is here.
Follow the steps below to use our project to backup your DynamoDB tables.
We've created a Serverless project on GitHub to create DynamoDB backups on a schedule. Usage of the project is simple.
First, install the Serverless Framework:
$ npm install -g serverless
You'll need AWS credentials configured in your terminal. Want help with these? Check out our walkthrough.
Then, use the Framework's install command to install a project template from a GitHub repo:
$ sls install --url https://github.com/alexdebrie/serverless-dynamodb-backups && cd serverless-dynamodb-backups
Edit the configuration in the
custom block of
serverless.yml to match your configuration. This includes setting the
tableName of your DynamoDB table, the
backupRate at which you want to create backups, the AWS region where your table is located, and optionally a
slackWebhook to send Slack notifications.
Want help setting up a Slack webhook? Check out the walkthrough.
Finally, deploy your Serverless service:
$ sls deploy Serverless: Packaging service... Serverless: Excluding development dependencies... Serverless: Creating Stack... Serverless: Checking Stack create progress... ..... Serverless: Stack create finished... Serverless: Uploading CloudFormation file to S3... Serverless: Uploading artifacts... Serverless: Uploading service .zip file to S3 (62.98 KB)... Serverless: Validating template... Serverless: Updating Stack... Serverless: Checking Stack update progress... ..................... Serverless: Stack update finished... Service Information service: serverless-dynamodb-backups stage: dev region: us-west-2 stack: serverless-dynamodb-backups-dev api keys: None endpoints: None functions: createBackup: serverless-dynamodb-backups-dev-createBackup
That's it - your service will create DynamoDB backups on your desired schedule! You're an Ops superhero.
If you want fancy Slack notifications when a backup succeeds or fails, follow the steps below.
In the end, you'll receive notifications like:
First, go to the channel you want to send notifications to and click Add an app:
In the page that opens, search for
Incoming Webhooks and click on it. Then click
Add Configuration. It should show your selected channel in the box. Then click
Add Incoming WebHooks Integration.
Once you've created it, the page will show your Webhook URL:
Copy and save this, as you'll need it in your Serverless service.
By default, it uses a boring
incoming-webhook display name and a boring image. I like to customize it a bit:
Paste your Webhook URL into the
serverless.yml as the
slackWebhook, deploy your service, and you'll start receiving notifications!
For the curious, I'll pass along some extra details and troubleshooting tips.
For some reason, not all DynamoDB tables are eligible to take a backup. I've found it's my older tables that don't allow backups.
If backups aren't enabled for your table, attempting a backup will throw a
When creating a backup, you need to specify a backup name. I was using the name of the table plus the ISO 8601 format. I kept getting an opaque error of:
Error: An error occurred (InternalServerError) when calling the CreateBackup operation (reached max retries: 9): Internal server error
I finally discovered that AWS doesn't allow colons in backup names. 🤔 Cryptic errors aside, I just changed my timestamp to be
Botocore uses a bunch of JSON files to describe the methods, inputs, outputs, and more of its various services. You can see the whole list here. To use a new operation, such as
create_backup() for DynamoDB, you need to make sure you have a version of
botocore with the proper models.
botocore are packaged into the AWS Lambda environment, which is nice most of the time. It means you don't have to package your own AWS API packages into your Lambda zip files. It's annoying in times like these, right after re:Invent, when the outdated
botocore dependency means you can use the newest methods.
Fortunately, you can upload your own
botocore data files without packaging your own version of
botocore! All you need to do is copy the data files for your desired models into your deployment package. You can see I've included the DynamoDB files here. Then, set the
AWS_DATA_PATH environment variable to the path where your data files are stored. I do it directly in my function handler before importing
import os # To get updated botocore data files os.environ['AWS_DATA_PATH'] = '.' import boto3 import botocore ...
Alex DeBrie is a data engineer at Serverless.
engineering-culture - 24.01.18
Our foray into mixing design thinking with agile development—what worked, and what didn't.
written by Brian Neisler
operations-and-observabilityguides-and-tutorials - 27.04.18
An end-to-end look at continuous integration with Python, Serverless and CircleCI.
written by Rupak Ganguly