mirror of
https://github.com/project-zot/zot.git
synced 2024-12-23 22:27:35 -05:00
33 lines
1.2 KiB
Markdown
33 lines
1.2 KiB
Markdown
|
# Profiling in Zot
|
||
|
|
|
||
|
|
This project gives the user the posibility to debug and profile the runtime to find relevant data such as CPU intensive function calls. An in-depth article on profiling in Go can be found [here](https://go.dev/blog/pprof).
|
||
|
|
|
||
|
|
A call to http://localhost:8080/v2/_zot/pprof/ would list the following available profiles, wrapped in an HTML file, with count values prior to change due to the runtime:
|
||
|
|
|
||
|
|
```
|
||
|
|
Types of profiles available:
|
||
|
|
Count Profile
|
||
|
|
95 allocs
|
||
|
|
0 block
|
||
|
|
0 cmdline
|
||
|
|
11 goroutine
|
||
|
|
95 heap
|
||
|
|
0 mutex
|
||
|
|
0 profile
|
||
|
|
13 threadcreate
|
||
|
|
0 trace
|
||
|
|
full goroutine stack dump
|
||
|
|
```
|
||
|
|
|
||
|
|
For example, the following can be used to gather the cpu profile for the amount of seconds specified as a query parameter, and then the results are stored in `cpu.prof` file:
|
||
|
|
```
|
||
|
|
curl -sK -v http://localhost:8080/v2/_zot/pprof/profile?seconds=30 > cpu.prof
|
||
|
|
```
|
||
|
|
|
||
|
|
Then, the user can use the `go tool pprof` to analyze the information generated previously in `cpu.prof`. The following command boots up an http server with a GUI and multiple charts that represent the data.
|
||
|
|
```
|
||
|
|
go tool pprof -http=:9090 cpu.prof
|
||
|
|
```
|
||
|
|
A flamegraph example would look like the following:
|
||
|
|
|
||
|
|
<img src="flamegraph.png" height="50%">
|