Another reason to avoid RDS

My list of reasons for never using or recommending Amazon’s MySQL RDS service grows every time I experience problems with customers. This was an interesting and still unresolved issue.

ERROR 126 (HY000): Incorrect key file for table '/rdsdbdata/tmp/#sql_5b7_1.MYI'; try to repair it

You may see this is a MyISAM table. The MySQL database is version 5.5, all InnoDB tables and is very small 100MB in total size.
What is happening is that MySQL is generating a temporary table, and this table is being written to disk. I am unable to change the code to improve the query causing this disk I/O.

What I can not understand and have no ability to diagnose is why this error occurs sometimes and generally when the database is under additional system load. With RDS you have no visibility of the server running the production database. While you have SQL access, an API for managing MySQL configuration options (I also add not all MySQL variables), and limited system statistics via a graphical interface, all information about the system performance, disk configuration etc is hidden and not accessible. This is a frustrating limitation of using RDS.

NOTE: While I cannot recommend RDS, I am very happy with AWS EC2 services when correctly configured. For a cloud based MySQL solution I would definitely recommend greater control over your MySQL database using EC2 and EBS.

Tagged with: Amazon Web Services (AWS) Databases EC2 MySQL RDS

Related Posts

My take on PyCon Australia 2026

PyCon AU 2026 held in Brisbane this week was a five day event for the Python community. With specialized tracks and dedicated tutorial and engineering days there was a wide variety of content available.

Read more

Fixing Old Software Bugs Without Thinking

I previously came across an issue in my benchmarking work where I relied on the Latency Histogram section of sysbench output. On Linux it worked fine — I generally ran my workload there, so it never mattered.

Read more

Curating DuckDB Datasets Leveraging AI

In Curated MySQL Data Sets for Realistic Testing I described datasets assembled the manual way — download, schema, load, validate, document — over hours or days per source. This post is the follow-up I promised: what changes when AI assists the same workflow, using DuckDB as the target engine and GeoNames as the first example.

Read more