memprofiler

Memprofiler helps to track memory allocations of your Go applications on
large time intervals. Go runtime implements multiple memory management
optimizations in order to achieve good performance, low heap allocation
cost and high degree of memory reuse. Therefore, sometimes it may be
tricky to distinguish “normal” runtime behaviour from real memory leak.
If you have doubts whether your Go service is leaking, you’re on the right
track. Memprofiler aims to be an open source equivalent of
stackimpact.com.

Getting started

Memprofiler is a client-server application. Memprofiler client is embedded
into your Go service and streams memory usage reports to the Memprofiler server.
Memprofiler server stores reports and performs some computations on the
data stream to turn it in a small set of aggregated metrics.
User will be able to interact with Memprofiler server via simple Web UI.

Client

To use Memprofiler in your application, run client in your main function:

package example
import (
    "time"
    "github.com/sirupsen/logrus"
    "github.com/memprofiler/memprofiler/client"
    "github.com/memprofiler/memprofiler/schema"
    "github.com/memprofiler/memprofiler/utils"
)
func main() {
    // prepare client configuration
    cfg := &client.Config{
        // server address
        ServerEndpoint: "localhost:46219",
        // description of your application instance
        ServiceDescription: &schema.ServiceDescription{
            ServiceType:     "test_application",
            ServiceInstance: "node_1",
        },
        // granularity
        Periodicity: &utils.Duration{Duration: time.Second},
        // logging setting
        Verbose: false,
    }
    // you can implement your own logger
    log := client.LoggerFromLogrus(logrus.New())
    // run profiler and stop it explicitly on exit
    profiler, err := client.NewProfiler(log, cfg)
    if err != nil {
        panic(err)
    }
    profiler.Start()
    defer profiler.Quit()
    // ...
}

Server

To run Memprofiler server, just install it and prepare server config
(you can refer to config example).

 ✗ GO111MODULE=on go get github.com/memprofiler/memprofiler
 ✗ memprofiler -c config.yml 
DEBU[0000] Starting storage                             
DEBU[0000] Starting metrics computer                    
INFO[0000] HTTP Frontend server resource                 URL=/schema.MemprofilerFrontend/GetSessions subsystem=frontend
INFO[0000] HTTP Frontend server resource                 URL=/schema.MemprofilerFrontend/GetServices subsystem=frontend
INFO[0000] HTTP Frontend server resource                 URL=/schema.MemprofilerFrontend/GetInstances subsystem=frontend
INFO[0000] HTTP Frontend server resource                 URL=/schema.MemprofilerFrontend/SubscribeForSession subsystem=frontend
INFO[0000] Starting service                              service=backend
INFO[0000] Starting service                              service=frontend