Aurora serverless (Postgresql) restarting several times per day

Question

Hi,

My Aurora DB in Serverless mode and on Postgresql engine worked fine in recent months. From begin new year (last two days) DB restarting several times per day with log note "DB instance restarted". DB is not available for few minutes after restart and works fine to next restart.

I also noticed that before each restart (~20s) is logged "The DB cluster has scaled from 8 capacity units to 4 capacity units, but scaling wasn't seamless for this reason: timed out finding a scaling point". Sometimes from 8 ACU to 4 ACU, sometimes from 4 ACU to 8 ACU.

Anybody know what is reason for this DB behavior?

Answer

When Serverless needs to scale, it looks for a  safe scaling point for 5 min, if it can not ( because of long running transactions) , then it either rolls back the scaling or if you have forced scaling enabled, it will kill current in-flight transactions and force a scaling. If it can not find a scaling point, then its quite possible that the current ACUs are unable to handle the new load and this can lead to a restart ( cpu brownout). So that could explain the restart when it goes from 4 to 8. however I see no reason for a restart during scale down. So its recommended that you should open a ticket with support.

Answer

I recommend that you create a support case for this issue.

Aurora serverless (Postgresql) restarting several times per day

相關內容