document how to analyze performance / memory usage issues

This commit is contained in:
Michael Eischer 2023-06-17 17:48:54 +02:00
parent a280b7364e
commit 3a32c4e59f

View file

@ -88,10 +88,40 @@ Then use the `go` tool to build restic:
$ ./restic version $ ./restic version
restic 0.14.0-dev (compiled manually) compiled with go1.19 on linux/amd64 restic 0.14.0-dev (compiled manually) compiled with go1.19 on linux/amd64
To create a debug build use:
$ go build -tags debug ./cmd/restic
You can run all tests with the following command: You can run all tests with the following command:
$ go test ./... $ go test ./...
Performance and Memory Usage Issues
===================================
Debug builds of restic support the `--block-profile`, `--cpu-profile`,
`--mem-profile`, and `--trace-profile` options which collect performance data
that later on can be analyzed using the go tools:
$ restic --cpu-profile . [...]
$ go tool pprof -http localhost:12345 cpu.pprof
To analyze a trace profile use `go tool trace -http=localhost:12345 trace.out`.
As the memory usage of restic changes over time, it may be useful to capture a
snapshot of the current heap. This is possible using then `--listen-profile`
option. Then while restic runs you can query and afterwards analyze the heap statistics.
$ restic --listen-profile localhost:12345 [...]
$ curl http://localhost:12345/debug/pprof/heap -o heap.pprof
$ go tool pprof -http localhost:12345 heap.pprof
Further useful tools are setting the environment variable `GODEBUG=gctrace=1`,
which provides information about garbage collector runs. For a graphical variant
combine this with gcvis.
Providing Patches Providing Patches
================= =================