A node.js microservices framework designs for scalability, simple to write and easy to maintain
Concept
Keep a microservice clean as much as posible and let Spinal do the mess part
We design Spinal
to help developers focus on the main objective of that microservice without to worry about other related systems like caching, queue, worker, load balancing. Spinal Nodes do the main task, anything else are left to a broker. Results are rapid development, more quality of microservice and easy to maintain in long term.
Installation
npm install spinal
Want some nigthly development trunk npm install jitta/spinal#development
Features
- Broker
- Broker will help all nodes connected
- Health check between nodes and remove fail nodes
- Load balance method calls from all nodes
- Hosted queue system
- Nodes
- Multiple namespace
- Call/Provide method from same namespace or other namespace (two ways)
- Cache result from method with specific hash key
- Provide worker to process jobs
Overview
Start
Spinal needs a broker to handle all call requests between namespace or nodes. Here is the code how to start a simple broker
var Broker = require('spinal').Broker;
var broker = new Broker();
broker.start(function(){
console.log('Spinal:Broker listening...' + this.port)
});
This code will start a broker at default port :7557
if you want a broker to listening on other port
broker.start(7777, function(){
console.log('Spinal:Broker listening on port 7777');
});
To add more broker features like queue system and caching system. You need to start a broker with a redis option.
var Broker = require('spinal').Broker;
var broker = new Broker({redis: 6379});
Call method
After we got a broker running, here is how to create a Spinal node and connect it to the broker that we have just started.
var Spinal = require('spinal').Node;
var spinal = new Spinal('spinal://127.0.0.1:7557', {
namespace: 'english'
});
spinal.provide('hello', function(data, res){
res.send('Hi ' + data.name);
});
spinal.start();
Now start another node to call the method
var Spinal = require('spinal').Node;
var spinal = new Spinal('spinal://127.0.0.1:7557', {
namespace: 'thai'
});
spinal.provide('hello', function(data, res){
res.send('Sawasdee ' + data.name);
});
spinal.call('english.hello', {name: 'hunt'}, function(err, result){
console.log(result); // Hi hunt
});
spinal.start();
Do not forget to start()
when you want to call some method. Use call()
and put namespace.method_name
Provide method
spinal.provide('name', function(data, res){
// send a result
res.send('A string');
res.send(12345);
res.send(true);
res.send({a: 1, b: 2});
// send an error
res.error('Error message');
res.error(new Error('Error message'));
// and support nodejs style callback
res(null , {a: 1, b:2})
res(new Error('Something wrong!'))
});
Connect a node to a broker
var spinal = new Spinal('spinal://127.0.0.1:7557', {
namespace: 'english', // assign node namespace
// OPTIONAL (in case run nodes and a broker on differrent machine)
// Specific host and port that we want this node to listen
// and want a broker to connect back to this node
hostname: '192.168.1.77',
port: 8888
});
Queue
To enable queue system we need a broker that start with redis.
// create a worker to process a job
// spinalA namespace is `newsletter`
spinalA.worker('send', function(data, res){
email.send(data.email, function(){
res.send('ok');
});
})
// create a job
spinalB.job('newsletter.send', {email: '[email protected]'})
.priority('high') // set priority: low, normal, medium, high, critical
.removeOnComplete(false) // if this is set to true, completed jobs (not failed) will be remove from Redis
.attempts(2) // use 2 times to process a job if it fail. try one more time
.ttl(10000) // timeout in 10s
.delay(5000) // before start delay for 5s
.backoff(10000) // after fail retry again in 10s
.onComplete(function(result) { console.log(result) }) // callback when job is done
.onFailed(function(err) { console.log(err) }) // callback when job failed
.save(function(err, job_id){ // don't forget to call save()
console.log('Created ' + job_id);
});
Call method with options
Timeout
Normally broker will set default timeout and return error to node if it's exceed 10 seconds but we can adjust it.
spinal.call('video.conversion',
{file: 'jitta.mp4'} // first argument need to be set
{timeout: 60000}, // set timeout option here!
function (err, result){
// if exceed timeout option will get an error
err.message === 'timeout error message'
}
)
Cache
Broker will cache result from the last method call if cache_id
present
in the options argument. It'll be hit cache after provide
cached data.
Note: All calls can use cache feature even a call inside the same namespace. To enable cache system we need a broker that start with redis.
spinal.provide('query', function(arg, res){
db.query('...', function(err, result)){
if(err) return res.error(err)
// cache for 1day with cache_id == sector+market
res.cache(3600, arg.sector + '-' + arg.market)
res.send(result)
})
})
spinal.call('stock.query',
{sector: 'Technology', market: 'US'},
{cache_id: 'technology-us'},
function (err, result, options){
options.from_cache === true // if it hit a cache
}
)
Automatically generate cache_id
from an arguments by set the cache_id
to true
spinal.provide('query', function(arg, res){
db.query(arg, function(err, result)){
if(err) return res.error(err)
// cache for 1day with cache_id == hashing(data)
res.cache(3600, true)
res.send(result)
})
})
spinal.call('stock.query',
{sector: 'Technology', market: 'US', jitta_score: {$gt: 8}}, {cache_id: true},
function (err, result, options){
options.from_cache === true // if it hit a cache
}
)
Testing
Sometime we want to make fixtures for all test case or stub a return result.
Spinal has a solution for this by capture all call
and replay in your test env
with spinal.nock.rec()
and spinal.nock.start()
var spinal = new Spinal('spinal://127.0.0.1:7557', {namespace: 'english'} );
spinal.nock('/path/to/fixtures') // target path that want to save fixtures
spinal.nock.rec() // enable record system
// and do normally like we do
spinal.start(function(){
spinal.call('email.send', {email: '[email protected]'}, function(err, result){
// then `err` and `result` will be save in the fixtures directory
// with filename `[email protected]`
})
})
Now it's time to replay the data that we saved.
var spinal = new Spinal('spinal://127.0.0.1:7557', {namespace: 'english'} );
spinal.nock('/path/to/fixtures') // send path to directory that we saved fixtures
spinal.nock.start() // then start nocking
spinal.nock.start({strict:false}) // or {strict:false} to by pass not exists fixture
spinal.start(function(){
spinal.call('email.send', {email: '[email protected]'}, function(err, result){
// `err` and `result` result will come from fixtures
})
})
If we need to stop recording or nocking by spinal.nock.stop()
Dashboard
Spinal comes with an internal dashboard to let us see what going on between
all microservices like numbers of nodes, methods, memory consumed, time usage
from each method, and load. This feature is provided via a broker so we need to
start a broker with more restapi
option.
var Broker = require('../').Broker;
var broker = new Broker({redis: 6379, restapi: 7577});
broker.start()
Then access localhost:7577
with your browser you will see it. Not just
a dashboard will start only. You will get queue dashboard (provide by
Kue) and some rest API in JSON format
/metrics - some useful metrics
/nodes - all nodes data
/methods - all methods
/queue/worker - number of workers
/queue/count - jobs count
Command Line
You can access spinal
command as a global by npm install -g spinal
in case
you might want to start broker easier spinal broker
or spinal broker -d
for
localhost devlopment enviroment. Incase you want to test a simple method
spinal call stock.query {sector:'Technology'}
Usage: spinal [options] [command]
Commands:
console run javascript console with spinal enviroment
call [options] <method> [data] call spinal method
job [options] <name> [data] create a job
broker [options] start a broker service
Options:
-h, --help output usage information
-V, --version output the version number
Roadmap
- Core
- Multi-Broker infinite Brokers to avoid bottlenecks and improve network reliability
- Event broadcast for each namespace (subscribe, emit)
- Message broadcast to all nodes
- Optimize performance
- Plugin System (put a plugin to run inside Spinal framework)
- Events Hook