How to Fix AWS Aurora Serverless Stuck in Modifying Mode After MinCapacity Change

Select Language:

If your Aurora Serverless v2 cluster seems to be stuck in a modifying state, it might be caused by a conflict that happened when you changed the minimum capacity to 0.5 ACU. Here’s a simple explanation and what you can do to fix it.

This issue happens because setting the minimum capacity to 0.5 ACU automatically limits the maximum number of connections, known as max_connections, to 2,000. Your cluster had this number set higher, at 5,000, which doesn’t match the new capacity setting. When AWS tried to adjust this, it could only change max_connections after restarting the cluster. Since the cluster was in the middle of a change, this caused it to hang.

The reason for this is that max_connections depends on the amount of memory allocated, which is based on the maximum ACU. But when you select a minimum capacity of 0.5 ACU, the system automatically sets max_connections to 2,000. Keep in mind, though, that having 2,000 connections with just 0.5 ACU doesn’t mean you can actually handle that many connections efficiently.

If your cluster is stuck during an update, your options are limited without support help. The static nature of max_connections means a restart is needed to change it, but since the cluster is already stuck, restarting may not be possible at this moment.

What you should do next is contact AWS support. They can help manually finish the update or roll it back if necessary. This is a technical issue that requires their intervention, especially since the cluster is not responding.

To avoid similar issues in the future, consider these tips:
– Use at least 1 ACU when expecting high connection loads.
– For many connections at lower ACUs, consider using a connection pooler like Amazon RDS Proxy or pg_bouncer to reduce memory strain.
– Always double-check parameter settings before changing capacity options to ensure they are compatible.

Since this problem appears to be a system-level hang, reaching out to AWS support is the best course of action. They can help resolve the issue and get your database running smoothly again.