In the current flow of DevOps practices, the "Workers" (also referred to as Runners, Agents, or Nodes) are the mysterious elements in charge of performing all of our builds, tests, and deployments. Prior to 2026, working with workers meant having to constantly update and scale them. Starting from 2026, the configuration of workers is about to become something completely different.
The following are some standards that a DevOps team needs to follow to make their lives easier starting from 2026.
1. The End of the "Long-Lived" Runner:
Among the changes that were made during the last couple of years, the most significant one has been a refusal from using pet runners in favor of ephemeral runners. Back before 2026, each DevOps team would run its runners on static hosts that had to operate for a long period of time, which led to the so-called "configuration drift."
Starting from 2026, all jobs should be performed in a new and clean environment. When finished with the job, this environment can be immediately destroyed.
2. AI-Enhanced Resource Right-sizing:
Should you allocate too much or too little CPU and RAM to your worker? Traditionally, this question was answered through estimation. Too much would be wasted; too little could result in failed builds or excessive build times.
With AI-optimized resource management, modern worker configurations use AI agents that analyze actual consumption of resources by your jobs. The AI agent detects that the "Frontend Build" runner uses only 20% of its allocated memory, then the system rightsizes the configuration for your next job execution without any additional effort from DevOps engineers.
3. FinOps: Cost-Oriented Worker Scheduling:
Uncontrolled infrastructure expenses may easily become an issue when working with unmanaged workers. In 2026, DevOps teams implement cost-aware scheduling through:
- Spot Instances Optimization: Non-critical processes such as linting and experimentation can be performed on "Spot" or "Interruptible" machines that are 90% cheaper than ordinary instances.
- Efficient caching: Using local dependency caches, workers need less time to download from the internet and more to execute code.
4. Security & Hardened Isolation:
With the evolution of attacks, the "Worker" has turned out to be one of the best targets for hackers seeking to infect the codebase with malicious code.
The latest standard requires Hardened Isolation. Not only do we require a basic container-based setup, but we also have to ensure that the jobs run in a MicroVM environment (like Firecracker). Moreover, each Worker will be provided with its own unique OIDC (OpenID Connect) Identity. As a result, the Worker can connect to any cloud services (S3 bucket, ECR repository), but it never needs to know the static API key/password.
5. Platform Engineering & "Golden Paths":
Instead of requiring our developers to dive into the depths of config.yaml, we've created so-called "Golden Paths."
By choosing one of the templates ("Standard Python Runner"), our engineers are provided with pre-configured options. All of the necessary configuration, patching, and scaling procedures are done in the background by our Platform Engineering team.
6. New Frontiers: Workers for Agentic AI:
The rise of Agentic AI creates an unknown approach to worker configuration. To perform a task, an AI agent must be able to run a piece of code that it writes itself. And here comes the need for the Ultrafast and ultra-isolated environment called Untrusted Code Hub, which can pop up in mere milliseconds, execute a piece of code, and then disappear. Agent Worker configuration should focus on low latency and Sandboxing technologies.
7. Implementing Worker Configuration the Right Way:
Here are three recommendations on how to bring your workers configuration to 2026 standards:
- Automate Cleanup: Make sure any temporary volumes or resource orphans do not outlive their usefulness within the job.
- Enforce Least Privilege: Never give a worker more permissions than they need for their specific task.
- AI-Driven Resource Intelligence: Stop guessing when assigning resources. Build in feedback mechanisms to have AI agents watch actual CPU/RAM usage data from your runners.
Worker configuration forms the core of your deployment engine. With ephemeral workers, agentic AI, and platform "Golden Path" strategies in place, your CI/CD becomes a powerful competitive asset.
To read more about How to Handle Employee Skill Management in Odoo 18, refer to our blog How to Handle Employee Skill Management in Odoo 18.