Profiling
Optimize Go applications using the built-in pprof package for CPU, memory, and goroutine profiling to identify bottlenecks and improve performance.
Profiling is an essential aspect of optimizing Go applications, as it helps identify bottlenecks, deadlocks, and inefficient code paths. Since the router is a Go application, you can leverage Go's built-in pprof
package for memory and CPU profiling. This section provides guidance on setting up and retrieving profiles using pprof
. These profiles are valuable for troubleshooting issues and can sometimes be the only way to gain meaningful context.
Enable Profiling
To enable the pprof
endpoints, start the router with the following environment variable:
The pprof
HTTP server will be accessible at http://localhost:6060
. Exposing this endpoint to production environments is highly discouraged due to security risks.
This makes the following endpoints available:
/debug/pprof/heap
— Memory profile./debug/pprof/profile
— CPU profile./debug/pprof/goroutine
— Goroutine profile./debug/pprof/threadcreate
— Thread creation profile./debug/pprof/block
— Block profile.
Downloading the Appropriate Profiles
To troubleshoot issues effectively, categorize them into the following three types:
CPU Utilization
Memory Utilization
Blocking and synchronization
Depending on the issue, you can download individual profiles or generate a ZIP archive containing a set of basic profiles. This is useful when you can't categorize the issue yourself.
1. CPU Utilization
To investigate CPU-related issues, you can fetch the CPU profile by running:
This command captures a 30-second CPU profile by default helping you identify functions consuming excessive CPU time.
2. Memory Utilization
To diagnose memory-related issues, you can download the heap profile:
This command captures a snapshot of memory allocations, allowing you to identify memory leaks or excessive memory usage.
3. Blocking and Synchronization
To identify deadlocks or goroutine-related issues, you can fetch the goroutine profile:
This command provides a detailed stack trace of all active goroutines, which is helpful for detecting deadlocks or excessive blocking.
Additional profiles for diagnosing blocking and synchronization issues:
Block Profile: Captures blocking events caused by synchronization primitives.
Thread Creation Profile: Identifies issues related to excessive thread creation.
By using these profiles effectively, you can pinpoint performance bottlenecks and improve the efficiency of your Go application. For further analysis, consider using the go tool pprof
interactive commands such as top
, list
, peek
, and web
.
Best Practices for Capturing and Sharing Performance Profiles
By attaching these profiles, you provide invaluable information for diagnosing performance bottlenecks and crashes efficiently! 🚀
Considerations Before Exporting:
Run with a real workload: Capture profiles during actual usage scenarios to ensure meaningful data.
Use an adequate duration: For CPU profiles, longer capture times (10–30 seconds) yield more useful information.
Avoid noise: Terminate other background processes to minimize interference in the profiles.
Automation Script for Linux and macOS
Use the following script to automate the steps described above, and then attach the archive to an issue or send it to us via Slack.
The resulting ZIP file will have the following flat structure:
Last updated
Was this helpful?