PerfectScale Introduces Automation for Argo Rollouts

Argo Rollouts are widely used by modern engineering teams to ensure smooth, non-disruptive delivery, enabling them to move fast and maintain a competitive edge. This is no longer a nice-to-have or exotic approach, but rather a commonly adopted and well-established practice. By enabling teams to deliver features gradually and ensuring they reach users safely, these strategies become crucial for eliminating risks associated with the deployment and releasing innovations with confidence.

Did the Argo rollout strategies become a standard? 

- YES!

Do Argo rollouts introduce a hidden layer of complexity for teams’ Kubernetes optimization policies? 

- YES - twice!!

Challenges with Argo rollouts

Gradual delivery offers clear benefits, but it also introduces complexity, with multiple concurrent versions of the same workload. This is how rollout works - yet it adds extra operational challenges related to Kubernetes workload resource allocation and optimization:

🚩 Multiple replicas of the same workload, from different versions, run simultaneously.

🚩 Imbalanced traffic distribution between different replicas of the same workload.

🚩 Resilience and stability can differ across different replicas.

🚩 Temporarily increased resource consumption due to parallel pipelines, etc.

All of this may result in misleading optimization signals, making traditional optimization efforts ineffective.

Here’s where PerfectScale by DoiT takes the stage! We’re excited to introduce our latest updates - advanced rollout-aware Automation is now available for PerfectScale customers!

Argo rollout-aware automation by PerfectScale

The new rollout-aware automation is a feature that finally aligns continuous optimization with continuous deployment. By understanding the full context of the implemented rollout strategy it enables teams to deliver faster without compromising K8s efficiency and stability. Here is how it works:

The Argo rollout strategy detection

PerfectScale automatically detects the implemented Argo rollout strategy, whether it is blue-green, canary, or A/B, without any manual configuration or tagging. 

Continuous per-replica analysis

With unique revision-awareness capabilities, PerfectScale can analyze and treat each replica of the workload independently. This approach enables teams to accurately evaluate the behaviour of each version, understand its performance and efficiency patterns, and define exactly how much resources it needs, unlocking safe and precise optimization across even the most advanced and diverse Kubernetes environments.

Autonomous rollout-aware optimization

PerfectScale provides granular control over automation actions, so it will always act accordingly. When the rollout support is enabled, PerfectScale autonomously applies right-sizing recommendations, taking into account each replica, aligning the optimization with your release approach.

When Argo rollouts break optimization, and how to fix it

Canary deployments

In a canary deployment, a small group of users gets the new version of the app, while the rest continue using the old one. In practice, the new version receives only 5-10 % of the traffic and then gradually rolls out, allowing engineers to test and verify it under different loads before fully replacing the old version. 

In that case, we are running into a situation where multiple revisions of the same workload are running simultaneously, each under a very different load. When it comes to optimization, the traditional tools treat these revisions as one workload, ignoring the fact of concurrent replicas, which may result in the canary replica, with a small fraction of traffic, wasting full-scale production resources.  

This is where PerfectScale becomes especially helpful. By continuously analyzing each replica independently, it right-sizes workloads based on their actual resource consumption, ensuring optimization actions reduce cost without introducing performance degradation or contradicting the rollout strategy.

Blue-Green deployments

When the Blue-Green rollout strategy is implemented, the system temporarily runs two full versions, old (blue) and new (green), until the green version passes all tests, and then switches the traffic completely from blue to green. As a result, multiple replicas run concurrently, adding operational complexity for K8s optimization and maintaining stability. 

In some cases, existing optimization solutions do their job, but due to the fact that they treat each version in the same way, this approach may be dangerous, leading to performance degradation and instability, or resource waste. 

This may happen when, due to configuration differences, the amount of memory that works well for the blue deployment is no longer enough for the green one. As a result, the green version may experience memory pressure, degraded performance, or even OOMs if both versions are provisioned with identical resources. On the opposite side, if the green version has been optimized and now needs less resources, applying the same configurations as the blue version leads to over-provisioning and wasted cloud spend.

To avoid these risks, PerfectScale seamlessly identifies the concurrent replicas, analyzes each one independently, and autonomously right-sizes the workloads based on the actual resource demand, enabling teams to optimize their K8s clusters safely while continuously driving value.

Reliable and cost-efficient optimization with Argo rollouts support

While advanced rollout strategies enable teams to do continuous delivery safely and reliably,  many organizations still postpone introducing them because of the operational overhead and additional infrastructure costs. PerfectScale support for Argo Rollouts helps teams bridge this gap and have reliable deployments without giving up on cost optimization.

📚 Want to see Argo rollouts support in action? Visit our documentation portal or book a technical session with our tech experts.

🆓 Still not with PerfectScale? Start for free today and enhance your K8s optimization journey.