Jan 28, 2022
Video: OtterTune “Human-in-the-Loop” mode
A few weeks ago, I wrote about OtterTune’s new “Human-in-the-Loop” (HITL) controls over automatic database configuration tuning. I recently presented a short talk and demo on the topic, and you can view the recording below.
The HITL mode enables OtterTune users to control what knobs OtterTune can automatically optimize and also control when and how changes are applied to the database. In the video, I explain the reasons for implementing HITL controls. We built them in response to users who requested the ability to apply change recommendations manually or who have DBMS change review/approval processes in place.
Then I demonstrated:
How to activate HITL controls
How to review new configuration change recommendations
Applying and skipping change recommendations
Controlling acceptable value ranges for configuration knobs
I also touched on additional HITL features we plan to add this year. The video is just 10 minutes long, and it’s a quick way to see how OtterTune let’s you manage automatic database optimization more closely:
Questions & Answers
After the talk, I received some interesting questions from the attendees, which I’m sharing below.
Q: Does using OtterTune result in downtime for my production databases? Does it have to restart the database for changes to take effect?
The answer depends on the knobs OtterTune changes. Some require a restart and some don’t. In OtterTune, you can actually set a switch that prohibits OtterTune from changing knobs that require restarts. We wrote a blog post related to this topic a few months ago.
Q: Can OtterTune run in my data centers?
As of January 2022, OtterTune can only tune Amazon RDS and Amazon Aurora databases (PostgreSQL and MySQL). We plan to add support for other cloud database platforms, other DBMSs, and eventually support for on-premises databases. If there are other DBMSs or platforms you’d like to see us support, just let us know in the OtterTune Slack channel.
Q: Does OtterTune also make recommendations for table partitioning, or disabling and creating indexes?
At this time, OtterTune recommends configuration tuning optimizations. We’ve done research into using machine learning to automate the optimization of database design and queries, etc., so that’s a possibility in the future.
If you have questions, I invite you to join the OtterTune community Slack channel and ask away!