Database Administrators Stack Exchange is a question and answer site for database professionals who wish to improve their database skills and learn from others in the community. It only takes a minute to sign up.
Sign up to join this community
I have a RDS instance that has workload at specific time window, however, cpu utilisation is a constant 20% even outside this window.
There is no replication configured.
Can't find a solution for this case.
Spec:
The database has more or less 2GB with a 300GB of storage (900 IOPS)
The instance type is a db.t2.medium
Quoting AWS support which I think was really disappointing:
"Thanks so much for your patience. I'm one of the RDS Support Engineers looking at your instance issue.
I've escalated this CPU utilisation issue of your instance to our RDS team and they are aware of an issue that affects a very small number of db.t2.medium instances where CPU spikes to 50% despite little to no actual database activity happening on your instance. This issue is outside of the database engine and in the underlying operating system. What this does is eat through the available CPU credits available to you, and when you get to zero CPU credits, your CPU starts being throttled, impacting the ability for your database to do any actual work. There's no remediation we can apply and we encourage you to scale your instance to either a db.t2.small or db.m3.medium or large which are not impacted by this issue. Our RDS team will be rolling out a proactive operating system fix to all impacted db.t2.medium instances over the coming months, although we don't have a timeline for this roll out.
Please let us know if we can help you further on this support case.
Best regards,
MJ A. | Sydney Support Centre Amazon Web Services We value your feedback. Please rate my response using the link below."
This is normal behavior from a t2.medium instance when it has burned off all its CPU credits. You'll notice on your graph that when CPU dies down a bit, the instance is suddenly capable of actually going to 100% for a limited period of time.
t2.medium instances can push 20% CPU per core as a baseline when all credits are burned based on documentation on the AWS website.
If you're constantly running into credit problems, it might be time to start looking at bigger instance types.
