Uploaded February 2024 | Updated September 2026, 2 weeks ago
Frederic Branczyk from Polar Signals shows us how to use Parca for identifying where memory was allocated and not yet freed in Go applications. In this particular case, we are diagnosing memory usage post-deployment, specifically in relation to recovering from a write-ahead log.
We see how to visualize memory allocation by functions and their children, with detailed examples of memory held by specific functions. The discussion extends to the recent support for heap profiling in Rust through a collaboration with Materialize, using the JEMalloc allocator.
We also touch on CPU profiling capabilities, distinguishing between kernel work and IO waiting times, illustrating this with kernel stack data. The ease of gathering this data using the Parca Agent, which requires no special configuration for collecting CPU profiling data, is also mentioned.
Listen to the audio part at 🎧 https://makeitwork.fm/1
The full content experience đź“ş makeitwork.tv
Frederic Branczyk from Polar Signals shows us how to use Parca for identifying where memory was allocated and not yet freed in Go applications. In this particular case, we are diagnosing memory usage post-deployment, specifically in relation to recovering from a write-ahead log.
We see how to visualize memory allocation by functions and their children, with detailed examples of memory held by specific functions. The discussion extends to the recent support for heap profiling in Rust through a collaboration with Materialize, using the JEMalloc allocator.
We also touch on CPU profiling capabilities, distinguishing between kernel work and IO waiting times, illustrating this with kernel stack data. The ease of gathering this data using the Parca Agent, which requires no special configuration for collecting CPU profiling data, is also mentioned.
Listen to the audio part at 🎧 https://makeitwork.fm/1
The full content experience đź“ş makeitwork.tv
