One of the more challenging aspects of optimizing Kubernetes is the rapid deployment of new code that continuously introduces changes to the environment. If new revisions are not optimized properly, it can lead to either wasted resources and budget or resilience issues that can impact your SLAs.
Understanding the release content and how it impacts your cost and resilience is a critical aspect of effectively keeping your environment optimized. However, getting the visibility into the data you need on your revisions is challenging, requiring complex configurations with observability tools, like Prometheus, that can often be error-prone.
To address these issues, and to provide our users with the most effective solution for keeping their K8s environment continuously optimized, we are excited to announce our latest capability: Revision Aware Optimization.
What is Revision Awareness?
PerfectScale’s Revision Awareness gives you a comprehensive breakdown of every workload revision that was released over the last 30 days. This level of visibility will allow you to easily compare versions and see when something happened, what was the issue, and if it was effectively remediated.
Revision Awareness is also a critical component of the PerfectScaler, which autonomously right-sizes your workloads. Vertical Pod Autoscaling (VPA)-based solutions, or any solution based solely on historical data, take actions based on summarized, over-time statistics. This can be problematic as it puts all the emphasis on what has happened in the past, but not taking into account that a developer may have made a resource change to support the future needs of their workloads.
The PerfectScaler is fully aware of your revision history and capable of detecting changes in CPU and Memory allocation patterns introduced by a new revision. This means automated actions are taken in a safe manner that is aligned with your application delivery process and supports the future growth of your application.
How to use Revision Awareness?
Revision Timeline View:
The timeline view helps you easily scan all the revisions for a given workload over the last 30-day period. The individual bars in the timeline represent a specific revision. Details on the revision are available by:
- Hovering: Providing you with a quick overview of resilience risks observed during the revision.
- Clicking: This will update the entire Zoom-in Window to provide more holistic details of the revision, including CPU and Memory Requests and Limits.
- Merging: This will allow you to merge every revision observed over the timeframe, giving you a clear picture of all the resiliency issues and wasted resources the workload experienced over the last month.
Note: The recommendations are displayed only for the current revision, which is based on the combination of historical utilization data with the current revision resource needs. Recommendations will not be available when viewing previous or merged revisions.
Revision History Details
Revision History details make it easy to scan every revision that has taken place over the last 30 days. This allows you to quickly see the deviations in CPU and Memory between each revision.
The Revision Types are defined as:
- New Revision: These are revisions that were deployed by your development team, or are changes based upon PerfectScale recommendations that were either manually added or triggered by an Alert or a Jira ticket.
- Automation: These are revisions that were initiated by PerfectScaler’s autonomous vertical scaling capabilities.
Other Updates: Improvement to the Zoom-In Window
With the addition of Revision Awareness, we took the opportunity to update the Zoom-In window to better align the workload details.
The major improvements are:
- Workload Details Panel: This panel contains all the essential data you need about the workload in an easy-to-scan format. Additionally, this panel will show you when you are looking at a workload that currently has the PerfectScaler Automations activated.
- Enhanced HPA Configuration Visibility: If a workload has HPA configured, not only are you able to see the maximum and average number of replicas, but now you can see the HPA trigger points for initiating autoscaling, broken down by either percentages or custom metrics.
- Save Resilience Level Configuration: You can now save the desired resiliency level at the workload level, which will overwrite the resiliency level at the cluster level. This gives you more flexibility in setting higher resiliency levels for mission-critical workloads while setting lower limits for workloads that require less headroom.
PrefectScale’s Revision Awareness is available on all of our subscription plans. To learn more about revision awareness visit our documentation portal, ask questions on the PerfectScale Community, or reach out to your account manager.
Not a PerfectScale customer, but want to learn more about Revision Awareness and the capabilities of our solution? Book a demo today!