Dave Cheney opined that there were only two kinds of logs needed: those used by production admin staff to keep the app running, and development debug logs. ^1
That seems correct; it just needs audit / accounting logs to cover the vast majority of use cases.
Second, production logs should be keyed, where the key 1) indicates the locality of the problem, and 2) is relatively persistent across releases. By default, slog will generate a key from the current source filename and function name for production logs.
This is for everyone who has googled log output, only to find that it was specific to only a few releases, and all the results were just the code itself.
The config has 4 options:
File = the filename for production or development logs
Debug = true for dev logs, false for prod
Prefix = keyword prefix for the locality hash
AuditFile = filename for audit logs -- optional
A barebones init:
slog.Init(slog.Config{
File: "STDERR",
Debug: false,
Prefix: "TEST",
})
slog.D("open file `%s'", cfg.File)
//...
slog.P("failed accessing `%s': %v", cfg.File, err)Which would yield the following because Config.Debug was false:
2017/07/19 21:07:44 WARN TEST-B0EAE04C failed accessing `/tmp/a': File not found
While a maximally verbose init:
slog.Init(slog.Config{
File: "STDERR",
Debug: true,
AuditFile: "/var/log/TEST.log",
Prefix: "TEST",
})
slog.D("open file `%s'", cfg.File)
//...
slog.P("failed accessing `%s': %v", cfg.File, err)
//...
slog.A("%s `%s' %s", r.Method, r.URL, r.RemoteAddr)2017/07/19 21:33:14.069128 go-log.go:53: open file `/tmp/a'
2017/07/19 21:33:14.069179 go-log.go:73: WARN TEST-B0EAE04C failed accessing `/tmp/a': File not found