Go logger is not well named. It's really an event queue that is managed by a hashset of event names.
logger.SetDefault(logger.NewAgentFromEnvironment()) // set the singleton to a environment configured default.
logger.Default().AddEventListener(logger.EventError, func(wr logger.Logger, ts logger.TimeSource, flag logger.EventFlag, args ...interface{}) {
//ping an external service?
//log something to the db?
//this action will be handled by a separate go-routine
})Then, elsewhere in our code:
logger.Default().Error(exception.New("this is an exception")) // this will write the error to stderr, but also
// will trigger the handler from before.You can defer writing a bunch of log messages to stdout to unblock requests in high-throughput scenarios. logger is very
careful to preserve timing state so that actions that live in the queue for multiple seconds are logged with the correct
timestamp.
You can standardize how you write log messages across multiple packages / services.
In _example/main.go you'll find a sample webserver with a bunch of endpoints, two worth looking at are /bench/logged and /bench/stdout. The former tests with the event queue driven logger, the later tests just writing to stdout.
Results:
Stdout / Printf:
> wrk -c 32 -t 4 -d 30 http://localhost:8888/bench/stdout
Running 30s test @ http://localhost:8888/bench/stdout
4 threads and 32 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 1.71ms 6.22ms 157.31ms 96.62%
Req/Sec 10.11k 1.11k 32.83k 88.84%
1208885 requests in 30.10s, 153.33MB read
Requests/sec: 40161.82
Transfer/sec: 5.09MBVersus event queue:
> wrk -c 32 -t 4 -d 30 http://localhost:8888/bench/logged
Running 30s test @ http://localhost:8888/bench/logged
4 threads and 32 connections
Thread Stats Avg Stdev Max +/- Stdev
Latency 1.37ms 8.53ms 151.83ms 99.02%
Req/Sec 12.80k 1.71k 34.19k 81.98%
1527807 requests in 30.10s, 193.79MB read
Requests/sec: 50758.87
Transfer/sec: 6.44MB
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent