I've been using a db.t4g.micro instance for my RDS, but I've been experiencing some crashes and significant slowdowns lately. I've just upgraded to a small instance, hoping this will help. Can anyone share advice on what metrics I should monitor to decide when to upgrade my RDS instance?
3 Answers
You should definitely keep an eye on the CPU and memory usage metrics. If those numbers look good, then check your storage metrics for any signs of IO-throttling. For GP3 volumes, make sure your IOPS stay below the provisioned amount. If you're using GP2, remember that your IOPS performance is tied to the volume size. Good luck!
It's crucial to monitor more than just one metric. A key one is query latency; if your queries are taking longer than usual, that's a red flag. The server should not crash unless there's a bug. From a systems perspective, start by checking memory usage and IOPS metrics. While monitoring memory is important, remember that databases cache a lot of data, so look also at your IOPS for cache thrashing. And keep an eye on CPU usage—especially on burst instances, as running out of CPU credits can really impact performance. If you're consistently hitting that limit, you might need to upgrade.
Cache thrash and memory issues can be spotted by checking the buffer pool hit ratio. A 100% hit ratio means your memory usage is efficient, but anything less indicates you're missing some data in memory and hitting the disk more often.
Make sure to track your burst credit usage. T instances build burst credits during low-traffic periods, and if you're constantly exceeding the baseline CPU usage, you'll start to drain those credits. I recommend upgrading whenever you consistently hit those usage levels. For crucial applications, I generally avoid T instances altogether.
Savior. Thank you!