Session recording crippling AWS Aurora Serverless database


I have a decently traffic’d site with about 10K visits per day, and am running Matomo on premise hosted on AWS ec2 (a1.xlarge 4 CPU, 8Gb ram) and database is an AWS Aurora serverless (at the moment fixed at 4 capacity units, 8 Gb ram).

I want to record 100% of sessions, daily, for 3 weeks, purging any sessions older than 3 weeks. However, whenever I create a recording, eventually the CPU usage of my RDS just continues to increase gradually until it is completely crippled. Normally at 4 capacity units (8 Gb ram) without session recording enabled, my RDS hums along barely being tested. When I turn session recording on resource consumption just continues to increase indefinitely until failure. I’ve even gone as high as 16 cap units (32Gb ram) and still the RDS is crippled eventually.

CPU usage spikes when session recording are on, and drops as soon as I turn them off…

The number of DB connections goes out of control in correlation with the above.

Does anybody have any idea why this could be happening?

Hi @c1dev - we have not seen this issue before, and we also run Matomo on AWS and our customers do as well. we can surely find the root cause and get this issue resolved.

I will reply by email to your message to troubleshoot further.

This issue was resolved after switching away from AWS Aurora serverless and back to a normal instance of AWS Aurora.

For some unknown reason, it appears as though AWS Aurora serverless and Matomo do not play nicely (at least in our case).