Can AI Help Automate IT Operations?

You've probably heard by now about how software developers can leverage AI to speed coding.

But what about ITOps engineers? Can they also use AI models to automate IT operations?

The answer appears to be yes, if my ITOps experiments with OpenAI are any indication. Using the text-davinci-003 AI model in the OpenAI playground, I tested how well AI could help ITOps teams perform common tasks, such as provision infrastructure and troubleshoot system failures.

Here's how OpenAI performed at the various tasks I presented to it in my endeavor to find out whether AI can help automate IT operations.

Configuring EC2 Instances with AI

The first test I performed represented a straightforward ITOps task: configuring cloud infrastructure using an infrastructure-as-code (IaC) approach.

I asked OpenAI to "Write a Terraform template to set up an EC2 instance." The result:

That's not perfect — this code would need some variable completion and tweaking to be used in the real world — but it's pretty good. If you're an ITOps engineer looking for a quick way to create an IaC template, it's easy to imagine AI being helpful.

IT Troubleshooting with AI

Next, I was curious how much OpenAI knows about common ITOps troubleshooting routines. So I posed a general question on this topic: "How can you troubleshoot a failed server?"

Here's what OpenAI said:

1. Check the server logs: Look in the server logs for any errors or warnings that may indicate why the server has failed.
2. Check the network connections: Check the network connections to make sure they are configured correctly and that there are no errors.
3. Check the hardware: Check the hardware components of the server to make sure they are all working properly.
4. Check the operating system: Check the operating system to make sure it is up to date and functioning properly.
5. Check the applications: Check the applications that are running on the server to make sure they are up to date and functioning properly.
6. Check the services: Check the services that are running on the server to make sure they are all running properly.
7. Check the power supply: Check the power supply to make sure it is providing the proper voltage and current.
8. Check the environment: Check the environmental conditions around the server to make sure it is not overheating.
9. Check the backups: Check the backups to make sure they are up to date and can be used to restore the server.

That advice is generic — it may or may not be useful in specific failure scenarios — but my question was also generic, so I can't really fault it.

Using AI for Log Monitoring

I wondered how well OpenAI would do if I posed a more focused question related to IT operations. So I asked, "What are the most important logs to monitor on an Ubuntu server?"

It gave a satisfyingly specific answer:

1. System Logs (syslog): This log provides a comprehensive record of system events such as system startup, shutdown, user logins, and application events.
2. Authentication Logs (auth.log): This log records all user authentication attempts, both successful and unsuccessful.
3. Security Logs (secure): This log contains information about system security events such as failed login attempts, file system access attempts, and system changes.
4. Application Logs (messages): This log contains information about application events such as errors, warnings, and status messages.
5. Apache Logs (access.log and error.log): These logs contain information about web server usage, errors, and requests.
6. MySQL Logs (mysql-error.log): This log contains information about any MySQL errors or warnings.

This is definitely not a perfect answer. I'm not sure what OpenAI means when it mentions "secure" and "messages" logs because Ubuntu doesn't have logs with those names. But the response is otherwise accurate enough for me to imagine this being useful for an ITOps engineer looking for a quick overview of the most important log files on Ubuntu Linux.

AI and Software Installation

Here's another response that I can imagine being useful.

When I asked OpenAI, "What is the command to reinstall a Debian package on Ubuntu?" it told me:

The command to reinstall a Debian package on Ubuntu is:
sudo apt-get install --reinstall [package_name]

That's totally right, and I can see how OpenAI would be useful for an engineer who needs a quick refresher on basic commands like this one.

Monitoring the Network with AI

Similarly, when I asked OpenAI to "Write a command to collect traffic with tcpdump," it responded:

tcpdump -i  -s 0 -w .pcap

If I were an ITOps engineer looking for the syntax to collect network traffic, it seems that OpenAI would be able to tell me what I need.

Conclusion

AI certainly isn't perfect at IT operations work. But it does better than you might expect at generating useful, mostly accurate information and code that ITOps engineers could use in the real world. And it will probably do even better if software vendors invest in AI tools designed specifically for ITOps — tools that don't yet exist but that could be on the horizon given all of the buzz at present surrounding AI tools for software development.

About the author

Christopher Tozzi is a technology analyst with subject matter expertise in cloud computing, application development, open source software, virtualization, containers and more. He also lectures at a major university in the Albany, New York, area. His book, “For Fun and Profit: A History of the Free and Open Source Software Revolution,” was published by MIT Press.

Comments

Plain text