profefe - Collect profiling data for long-term analysis

profefe continuously collects profiling data from a running Go application and provides an API for querying the profiling samples base on metadata associated with the application.

The project is still in the early prototyping stage. Things will change. Opinions and contributions are welcome.

Why Continuous Profiling?

"Continuous Profiling and Go" describes the motivation behind the project.

How does it work?

See Design Docs documentation.

Quickstart

TODO add quickstart

To build and start profefe collector, run:

> make
> ./BUILD/profefe -addr :10100 -log.level debug -badger.dir /tmp/profefe

2019-06-06T00:07:58.499+0200    info    profefe/main.go:86    server is running    {"addr": ":10100"}

The project includes a fork of Google Stackdriver Profiler's example application, modified to use profefe agent, that sends profiles to the local collector.

To start the example, in a separate terminal window run:

> go run ./examples/hotapp/main.go

After a brief period, the application will start sending CPU profiles to the collector:

send profile: http://localhost:10100/api/0/profiles?instance_id=87cdc549c84507f24944793b1ddbdc34&labels=version%3D1.0.0&service=hotapp-service&type=cpu
send profile: http://localhost:10100/api/0/profiles?instance_id=87cdc549c84507f24944793b1ddbdc34&labels=version%3D1.0.0&service=hotapp-service&type=cpu
send profile: http://localhost:10100/api/0/profiles?instance_id=87cdc549c84507f24944793b1ddbdc34&labels=version%3D1.0.0&service=hotapp-service&type=cpu

Querying Profiles

Querying profiling data is an HTTP call to profefe collector API endpoint:

> go tool pprof 'http://localhost:10100/api/0/profiles/merge?service=hotapp-service&type=cpu&from=2019-05-30T11:49:00&to=2019-05-30T12:49:00&labels=version=1.0.0'

Fetching profile over HTTP from http://localhost:10100/api/0/profiles...
Saved profile in /Users/varankinv/pprof/pprof.samples.cpu.001.pb.gz
Type: cpu
Entering interactive mode (type "help" for commands, "o" for options)
(pprof) top
Showing nodes accounting for 43080ms, 99.15% of 43450ms total
Dropped 53 nodes (cum <= 217.25ms)
Showing top 10 nodes out of 12
      flat  flat%   sum%        cum   cum%
   42220ms 97.17% 97.17%    42220ms 97.17%  main.load
     860ms  1.98% 99.15%      860ms  1.98%  runtime.nanotime
         0     0% 99.15%    21050ms 48.45%  main.bar
         0     0% 99.15%    21170ms 48.72%  main.baz
         0     0% 99.15%    42250ms 97.24%  main.busyloop
         0     0% 99.15%    21010ms 48.35%  main.foo1
         0     0% 99.15%    21240ms 48.88%  main.foo2
         0     0% 99.15%    42250ms 97.24%  main.main
         0     0% 99.15%    42250ms 97.24%  runtime.main
         0     0% 99.15%     1020ms  2.35%  runtime.mstart

Note, above we requested all profiling data associated with the given meta data (service and time period), as a single merged profile.

profefe includes an experimental tool, that allows importing existing pprof data into the collector. While the collector running, run the tool as following:

> ./scripts/pprof_import.sh --service service1 --label region=europe-west3 --label host=backend1 --type cpu -- path/to/cpu.prof

uploading service1-cpu-backend1-20190313-0948Z.prof...OK

HTTP API

Save pprof data

POST /api/0/profiles?service=<service>&instance_id=<iid>&type=[cpu|heap]&labels=<key=value,key=value>
body pprof.pb.gz

service — service name (string)
instance_id — an identifier of running instance (string) (TODO: why do we need instance_id?)
type — profile type (cpu, heap, block, mutex, or goroutine)
labels — a set of key-value pairs, e.g. "region=europe-west3,dc=fra,ip=1.2.3.4,version=1.0" (Optional)

Query saved meta information

GET /api/0/profiles?service=<service>&type=[cpu|heap]&from=<created_from>&to=<created_to>&labels=<key=value,key=value>

service — service name
type — profile type
from, to — a time window between which pprof data was collected
labels — a set of key-value pairs

Query saved pprof data returning it as a single merged profile

GET /api/0/profiles/merge?service=<service>&type=[cpu|heap]&from=<created_from>&to=<created_to>&labels=<key=value,key=value>

Request parameters are the same as for querying meta information.

Return individual pprof data

GET /api/0/profiles/<id>

id - id of stored pprof file; returned with the request for meta information query

Feedback

The feedback and contribution are very welcome.

License

MIT

akovardin / profefe