Recently I observed some database connection errors in my (mostly idle) application. Monitoring showed that my db cpu credit balance had gradually decreased to 0 starting a few weeks ago, when the instance's CPU utilization crept above 10% (it's a t2.micro). This was constant usage, despite the only application (/EC2 instance) that uses the database being idle. The application was noticeably slower than normal, despite monitoring on the EC2 showing no issues. I tried rebooting both database and EC2 instance - since then (~1 day later) the database CPU credits have remained at zero, although the "CPUSurplusCreditBalance" is slowly increasing, the CPU utilisation remains slightly above 10%.
I have enabled enhanced monitoring and performance insights. Post-restart, they don't show any activity beyond RDS-controlled queries (and those aren't frequent). Furthermore, the OS process list shows "RDS processes" as the highest contributor to CPU usage, although that's only around 4% (I guess it's hard to reconcile a snapshot of processes against overall CPU% usage). There are no open or idle connections to the database.
Looking at monitoring and logs from when the credit drainage first started (April 10th), I don't see anything special happening at all (performance insights doesn't go back that far).
My case seems somewhat similar to this one from 2015, though this is now occuring post-reboot (i.e. it's not caused by a single burst of activity). https://dba.stackexchange.com/a/118939
Rather frustrating that despite being a paying customer, I am unable to open a technical support case. Hoping this forum can help!
AWS 官方已更新 3 年前