The memory consumption for "RDS Processes" is slowly but steadily increasing, even though there is no activity on the instance. There are no connections to the database, and using the RDS Console's OS Process List view I can see that memory usage for all processes remains steady or declines over time except for one. I created a SQL Server Express instance with Enhanced Monitoring enabled. I am running an experiment on a t2.micro instance as we speak, and while the results will be more obvious in a few hours I believe I see the source of the problem. The problem may be t2.micro specific, but more likely the t2.micro is such a memory constrained instance type that any change in memory usage impacts its behavior.
Probably what you are seeing is that RESOURCE MONITOR is constantly trying to free memory because something external to sqlservr.exe is using increased amounts of memory and RESOURCE MONITOR run frequently to adjust SQL Server's memory use as a result. Since this has been known about, but not "fixed", for many years it is likely not a bug. SQL Server's RESOURCE MONITOR process is known (based on web searches) to utilize a lot of CPU when SQL Server is under memory pressure. AWS should resolve this in the next few days, but in the meantime I'm going to suggest workarounds. So despite there being no changes to the database schema, application code, or applied load this jumps to the top as a culprit. Changes to the data will eventually trigger SQL Server to automatically re-gather statistics and recompile plans, and you could get one that is far more CPU-intensive than a previous plan. This is completely unrelated to RDS, and is often the result of a change to a query plan. Sudden increases in CPU utilization, including going to 100%, has frequently been associated with SQL Server (and other relational database systems, BTW) throughout history.
Even there the reason would most likely be that the SQL Server update caused a query plan to change. It still could be related to an update being applied to SQL Server (or less likely Windows), so if you were patched recently then that is suspicious too.
But if it is a SQL Server-associated process it most likely is not. So if you found, via Enhanced Monitoring for example, that a process not normally associated with SQL Server was eating up all the CPU then perhaps you've identified an RDS issue. Also, there is a management agent that is running on the system.
It does install stored procedures and triggers (e.g., to capture and block DDL that might damage automation), but can't change the source code.
While you can never rule out a problem anywhere without investigation, RDS itself does (and can do) little to SQL Server itself. Indeed the reason that technical support isn’t free (and obviously if it was they’d have to raise instance prices to cover it) is that 95%+ (I’d guess) of the time people are looking for help it really does not turn out to be an AWS issue. Unfortunately Basic Support (the free plan) does not include technical support and you only get it here in the forum sporadically (when someone on the RDS team looks in and is curious). They could then report back the findings, and others could determine if the same might apply to them.
In the end it looks like someone is going to have to open a technical support case to have their instance looked at. What’s going on with your burst credits? There are so many things to look at, and it isn’t clear that elevated CPU utilization has anything to do with RDS specifically. Have you used Enhanced Monitoring to look at which processes are using the CPU? Have you tried the SP_WhoisActive or other tools for examining SQL Server activity? (SP_WhoisActive can be downloaded from but requires some tweaking to work on RDS, which you can find via web searches). These may or may not be related incidents.