This post will talk about backup and disaster recovery strategy on AWS RDS been made by lambdas. First of all let’s talk about the default snapshot that comes enable on every RDS and from my sight is a really good solution with its point-in-time option that takes log backups each every 5 minutes which allow us huge coverage on failures. There’s 2 types of snapshot automated and manual, automated are taken daily and incremented all day long to delivery the point-in-time solution these snapshot are holding for 35 days max, and this isn’t configurable, the manual are made manually, through the CLI or by application and there’s no expire date and are limited to 100 per instance and region but you can increase this number opening a support ticket with AWS. An important detail here is the possibility to take copy of these snapshots for both but from automated the result always became a manual, and this will be our strategy today.

Depending on which is you business and legal implications you have to keep copies for you data for amount of time. In this case automated copies from the existing backups is pretty handful, although is interesting put a a limit date to keep all these backups. For this reason we have 2 pair for lambda one for made the backup and another to remove it. When is comes to disaster recovery I won’t present a full action automated plan to take a minimum downtime possible today, instead will show the base action of do the backups and copy in a different region in case of failure of the main one. Remind yourself that we are taking care only of your RDS snapshot for now, to a full disaster recovery your application has to be copied as well as all the dependencies, and if you want to come backup to the main region all the steps until now have to made in the opposite direction.

After this brief explanation let go to the code, although I’m a DBA still have to dominate python so this time will be javascript other tools would be AWS CLI configured, nodejs >=10 with serverless framework installed , this framework is used to deploy on AWS. Full version of this code is on github

AWS SDK is huge so I got just the function I need on the applications in a single file rds-func.js

First detail that’s pops-up is that cluster and instances are treat by different call on SDK following by its snapshots

Another detail is to search snapshots you have to know the resource’s names (cluster or instance), because of that all lambdas perform a query to list all resources for both copy or delete.

Below we have a list of all daily cluster backups, checked by date from a function that will remove days from today and right after will call copy function. As all calls are asynchronous after lambda execution you can follow the copy through the console, depend of the database size of course. On the second piece all this process is repeated to instances.

Delete follow the same structure but instead of copy function you have delete

On the function that copy to another region comes a interesting detail, the search has to be realized from the original region but the copy will be executed from the destiny region and informed the original region. Another detail here is that copy function will reference the snapshot by ARN instead the name.

Delete function is the same but one more time just replacing copy for delete

If you look inside the function all variables comes from environment variables that were declared on the mail file serverless.yml , this file is used by the framework. Below we inform the provider, the region where we will deploy you local aws profile and an IAM with RDS access (you can change for one more restrict one) and the variables

daysBefore: keepDays & keepLongDays:

mainRegion & copyRegion:

The packaging can be customized in that case I just use it to exclude some files

Below we finally have the function declaration in this order name, file + function name, timeout and a event trigger in our case is a cron schedule . I prefer this approach of deploy all function at once but it’s possible deploy one function at the time, you will see right below.

Ok, everything is set now is the time to send all to the cloud. This code below is all you need

serverless deploy

Or function by function

serverless deploy function --function manual-backup

To remove

serverless remove

Here we can say that we can sleep more peacefully knowing that our RDS are protected from possible failures. The basics of snapshots and availability in other regions are covered here, it is obvious that a disaster recovery solution for your environment will be more complete and will involve more technologies and other forms of automation.

There’s a version in Portuguese here

var i = snapName.indexOf(“snapshot:rds”);
var tmpName = snapName.substring(i, snapName.length)
return tmpName.replace(‘snapshot:’,’’).replace(“rds:”, “manual-”);
return snapName.replace(“rds:”, “manual-”)
}module.exports = {

Musician, DBA, developer, tech enthusiast and maybe devops

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store