Query on Beyla's Compatibility and Deployment for Large Infrastructures

[ANSANJAY]

Hello Beyla Team,

I'm currently evaluating various observability solutions to identify the best fit for our infrastructure needs, and Beyla has caught our attention due to its promising capabilities. Before proceeding further, I have a few questions regarding Beyla's support and deployment flexibility, particularly in complex environments:

1. Kubernetes Support: Could you clarify if Beyla offers out-of-the-box support for Kubernetes clusters? We're looking for solutions that can seamlessly integrate into our K8s environment.
2. VM Monitoring: Is Beyla capable of monitoring virtual machines as well? This capability is crucial for us as our infrastructure spans both containerized and traditional VM-based applications.
3. Customization in VM Environments: For VM deployments, does Beyla allow for any level of customization to cater to specific monitoring needs or challenges?
4. Ideal Deployment Scenarios: In your view, what are the scenarios or environments where Beyla truly excels? Understanding the ideal use cases would help us gauge how well it aligns with our infrastructure.
5. Deployment Strategies for Large Fleets: Lastly, could you offer any insights or recommendations on deploying Beyla across a significant number of VMs and Kubernetes clusters? If there are examples of large organizations utilizing Beyla, that would also be helpful to know.

We appreciate the work you've put into developing Beyla and look forward to potentially integrating it into our observability stack. Thank you for addressing these queries.

Best regards,
Anamika

[grcevski]

Hi Anamika, I'll answer your questions in order, but if I've missed something, I apologize in advance and please let me know.

1. Kubernetes Support: Could you clarify if Beyla offers out-of-the-box support for Kubernetes clusters? We're looking for solutions that can seamlessly integrate into our K8s environment.

Yes. We support automatic gathering of k8s attributes without any external agent requirements. We also support choosing which pods to instrument by defining Kubernetes pod pattern matches. You can check-out our Kubernetes tutorial here:
https://grafana.com/docs/beyla/next/tutorial/k8s-walkthrough/

2. VM Monitoring: Is Beyla capable of monitoring virtual machines as well? This capability is crucial for us as our infrastructure spans both containerized and traditional VM-based applications.

Yes. Beyla work's on VMs, but we currently don't automatically capture the cloud vendor tags. It's in plan for us to add support for AWS, GCP and Azue. For now, you'd have to manually add the attributes you need for the VM's by using OTEL_RESOURCE_ATTRIBUTES, which will get picked up by Beyla.

3. Customization in VM Environments: For VM deployments, does Beyla allow for any level of customization to cater to specific monitoring needs or challenges?

Can you please elaborate what you mean by this? We do have extensive configuration options to support various scenarios, but I'd need more information on the types of challenges you are facing to see if we allow those customizations.

4. Ideal Deployment Scenarios: In your view, what are the scenarios or environments where Beyla truly excels? Understanding the ideal use cases would help us gauge how well it aligns with our infrastructure.

I think Beyla's main strength is that it will immediately help you get Application Observability signals across number of programming languages, so it really excels when you have large and diverse set of applications, especially if some of those use libraries which aren't supported by the OpenTelemetry SDKs and agents.

5. Deployment Strategies for Large Fleets: Lastly, could you offer any insights or recommendations on deploying Beyla across a significant number of VMs and Kubernetes clusters? If there are examples of large organizations utilizing Beyla, that would also be helpful to know.

We've seen couple of approaches so far:

• On Kubernetes some organizations deploy one Beyla as a DaemonSet per node, which instruments as pods come up and down.
• Other organizations like to use sidecars on Kubernetes, so they use our Kubernetes operator https://github.com/grafana/beyla-operator.
• For VMs, I think the best strategy is to build Beyla as part of your image and have it start as the VM boots. A good starting point here would be our systemd example https://github.com/grafana/beyla/tree/main/contrib.

We have a few large organizations using Beyla, unfortunately I cannot mention any names.

If you'd like to ask more of these questions perhaps though chat or face to face, you can join our community #beyla slack or we also have our first community call on February 14th https://calendar.google.com/calendar/u/0/embed?src=grafana.com_n57lluqpn4h4edroeje6199o00@group.calendar.google.com.