Adding Observability in CI and System Testing
In modern embedded systems development, integrating observability into Continuous Integration (CI) and system testing is crucial for ensuring software reliability and performance.
Moreover, observability can speed up overall product development by faster issue resolution, in particular by reducing the need to reproduce reported issues in a separate debugging environment.
This talk will explore techniques for early detection of system anomalies, including random crashes and latent stability risks.
We will discuss methods for collecting debugging data like call-stack traces, system traces, and logs for effective diagnostics on detected anomalies.
Additionally, strategies for managing data flow from devices to development teams will be covered, emphasizing seamless integration and collaborative workflows.
Attendees will gain insights into implementing proactive observability practices to enhance software quality and accelerate development cycles.
What is the primary difference between 'observability' and classic 'debugging' as described in the talk?
Time on X axis: Yes, you can change to horizontal mode in the View menu (the local menu in the Trace view).
Instrumentation: For tracing of supported RTOSes like FreeRTOS and Zephyr, the instrumentation for kernel events like taskswitches and API calls is already provided by TraceRecorder.
Percepio Detect: Yes, the Server, Client and Receiver can all run on the same computer. Thats the default setup in the bundled demo.
I
Ah that's good. I would have liked an affordable option for personal learning use (the academic licence doesn't apply). At work we use SystemView, but I often find it lacking.
We offer a free version via certain RTOS partners, and for the paid version we are pretty flexible when it comes to licenses for non-commercial use. Contact me to learn more (johan.kraft@insert-company-here.com).
Hi
Interesting product you push at the end ;-)
How big is such a trace packet?
I know it depends, but assume i log the same info as you show in the demo. And just in decades (10^?).
Regards
In the demo I showed, the "alert" packets are about 7-8 KB if I recall correctly. Most of this is the 6 KB trace buffer. With only core dumps included, the full alert packets are just under 1 KB in this demo, but can be made a bit smaller if needed.
Excellent presentation!
Thanks!
Can Tracealyzer display with time on the x axis instead? I had a demo copy once, but it felt disorientating with the vertical view and there were so many options. How much instrumentation of the code is needed to get going? I really like the simplicity of the Percepio Direct dashboard. Can all 3 computers be one and the same, i.e. the board connects directly to my PC USB port to help debug?