Strange DB CPU Utilization in RDS Aurora Postgres Writer instance

[Jose_Miguel_Concha]

Hi

We are running Chirpstack on a HA/AS architecture in AWS. Last week we have experimented a strange DB activity that get normalized after an instance reboot.

We are using a RDS-Aurora Postgres cluster with one reader and one writer instance.

The strange activity was on the writer instance. And here are the logs from those days. After the restart the logs stopped recording that activity
Reviewing the documentation I could not find a possible cause of the problem to take the corresponding actions.

Thanks in advance.

Regards,

Link to the logs (public access):

postgresql.log.2024-07-12-1800

[Jose_Miguel_Concha]

For some reason the shared memory started to rise till I rebooted the DB instance. I have terminated the EC2 instances too to see if there was something strange from the app side. Now the buffer cache hit ratio get lower and that means that the I/O requests in the DB are getting to the clouds, producing a higher cost.