Check out our main project at safing/portmaster
Portbase
Portbase helps you quickly take off with your project. It gives you all the basic needs you would have for a service (not tool!).
Here is what is included:
log
: really fast and beautiful logging
modules
: a multi stage, dependency aware boot process for your software, also manages tasks
config
: simple, live updating and extremely fast configuration storage
info
: easily tag your builds with versions, commit hashes, and so on
formats
: some handy data encoding libs
rng
: a feedable CSPRNG for great randomness
database
: intelligent and syncable database with hooks and easy integration with structs, uses buckets with different backends
api
: a websocket interface to the database, can be extended with custom http handlers
Before you continue, a word about this project. It was created to hold the base code for both Portmaster and Gate17. This is also what it will be developed for. If you have a great idea on how to improve portbase, please, by all means, raise an issue and tell us about it, but please also don't be surprised or offended if we ask you to create your own fork to do what you need. Portbase isn't for everyone, it's quite specific to our needs, but we decided to make it easily available to others.
Portbase is actively maintained, please raise issues.
log
The main goal of this logging package is to be as fast as possible. Logs are sent to a channel only with minimal processing beforehand, so that the service can continue with the important work and write the logs later.
Second, is beauty, both in form what information is provided and how.
You can use flags to change the log level on a source file basis.
modules requires log
packages may register themselves as modules, to take part in the multi stage boot and coordinated shutdown.
Registering only requires a name/key and the prep()
, start()
and stop()
functions.
This is how modules are booted:
init()
available: flags, config, logging, dependencies
- register flags (with the stdlib
flag
library)
- register module
module.prep()
available: flags, config, logging, dependencies
- react to flags
- register config variables
- if an error occurs, return it
- return ErrCleanExit for a clean, successful exit. (eg. you only printed a version)
module.start()
available: flags, config, logging, dependencies
- start tasks and workers
- do not log errors while starting, but return them
module.stop()
available: flags, config, logging, dependencies
- stop all work (ie. goroutines)
- do not log errors while stopping, but return them
You can start tasks and workers from your module that are then integrated into the module system and will allow for insights and better control of them in the future.
config requires log
The config package stores the configuration in json strings. This may sound a bit weird, but it's very practical.
There are three layers of configuration - in order of priority: user configuration, default configuration and the fallback values supplied when registering a config variable.
When using config variables, you get a function that checks if your config variable is still up to date every time. If it did not change, it's extremely fast. But if it, it will fetch the current value, which takes a short while, but does not happen often.
// This is how you would get a string config variable function.
myVar := GetAsString("my_config_var", "default")
// You then use myVar() directly every time, except when you must guarantee the same value between two calls
if myVar() != "default" {
log.Infof("my_config_var is set to %s", myVar())
}
// no error handling needed! :)
WARNING: While these config variable functions are extremely fast, they are NOT thread/goroutine safe! (Use the Concurrent
wrapper for that!)
info
Info provides a easy way to store your version and build information within the binary. If you use the build
script to build the program, it will automatically set build information so that you can easily find out when and from which commit a binary was built.
The build
script extracts information from the host and the git repo and then calls go build
with some additional arguments.
This is just a convenience wrapper around encoding/binary
, because we use varints a lot.
DSD stands for dynamically structured data. In short, this a generic packer that reacts to the supplied data type.
- structs are usually json encoded
- []bytes and strings stay the same
This makes it easier / more efficient to store different data types in a k/v data storage.
rng requires log
, config
This package provides a CSPRNG based on the Fortuna CSPRNG, devised by Bruce Schneier and Niels Ferguson. Implemented by Jochen Voss, published on Github.
Only the Generator is used from the fortuna
package. The feeding system implemented here is configurable and is focused with efficiency in mind.
While you can feed the RNG yourself, it has two feeders by default:
- It starts with a seed from
crypto/rand
and periodically reseeds from there
- A really simple tickfeeder which extracts entropy from the internal go scheduler using goroutines and is meant to be used under load.
database requires log
introduction to be written
api requires log
, database
, config
introduction to be written
The main program
If you build everything with modules, your main program should be similar to this - just use an empty import for the modules you need:
import (
"os"
"os/signal"
"syscall"
"github.com/safing/portbase/info"
"github.com/safing/portbase/log"
"github.com/safing/portbase/modules"
// include packages here
_ "path/to/my/custom/module"
)
func main() {
// Set Info
info.Set("MySoftware", "1.0.0")
// Start
err := modules.Start()
if err != nil {
if err == modules.ErrCleanExit {
os.Exit(0)
} else {
os.Exit(1)
}
}
// Shutdown
// catch interrupt for clean shutdown
signalCh := make(chan os.Signal)
signal.Notify(
signalCh,
os.Interrupt,
syscall.SIGHUP,
syscall.SIGINT,
syscall.SIGTERM,
syscall.SIGQUIT,
)
select {
case <-signalCh:
log.Warning("main: program was interrupted")
modules.Shutdown()
case <-modules.ShuttingDown():
}
}