AWS RDS Multi-AZ “Storage Full” Failover

Question

Quick Question,

Most probably know the answer, but if using Multi-AZ RDS with MySql and the primary storage is full, then the standby replica will also be full as well? Failover would only be successful in this situation should the primary storage just have an issue, but not necessarily be full.

What alternatives would you use in this case besides Cloudwatch and RDS metric for storage?

Answer 1

Yes, you are correct regarding the standby storage space.

You should have your CloudWatch alerts configured to notify you well ahead of time. For example you could configure it to start sending you emails when the storage is at 80% full, and make plans to increase your storage space well before it hits 100%.

I'm not sure what alternatives to CloudWatch/RDS metrics you would need, or what you are looking for in an alternative for RDS storage metrics. What specifically is CloudWatch/RDS not providing you that you feel you need?

Answer 2

The alternative is to use Amazon Aurora.

They promise that storage grows dynamically as you use it. You pay only for the storage you use. It grows as much as S3 can grow, which is greater than you'd want any MySQL instance to grow.

I still have some questions about Aurora—like what happens if you expand your InnoDB tablespace, and can't shrink it? Do you have to dump and restore everything? (I'm assuming we've disabled file-per-table.)

I suppose one could create a slave manually and then switch over to the slave with minimal downtime.