Proxy
BootcampFrontend2020-12-17
1.request
var request = require( 'request' );
app.all( '/proxy/*', function( req, res ){
req.pipe( request({
url: config.backendUrl + req.params[0],
qs: req.query,
method: req.method
})).pipe( res );
});It works fine if remote host is available, but if remote host is unavailable the whole node server crashes with unhandled exception
app.all( '/proxy/*', function( req, res ){
req.pipe( request({
url: config.backendUrl + req.params[0],
qs: req.query,
method: req.method
}, function(error, response, body){
if (error.code === 'ECONNREFUSED'){
console.error('Refused connection');
} else {
throw error;
}
})).pipe( res );
});You should actually try to prevent the ECONNREFUSED exception from becoming uncaught:
var request = require( 'request' );
app.all( '/proxy/*', function( req, res ){
req.pipe( request({
url: config.backendUrl + req.params[0],
qs: req.query,
method: req.method
}))
.on('error', err => {
const msg = 'Error on connecting to the webservice.';
console.error(msg, err);
res.status(500).send(msg);
})
.pipe( res );
});If you get an actual uncaught exception, then you should just let the application die.
- If you store any state on your servers that you don’t want to loose when a server is down then store your state on the two specific database servers - a master and a replica. If you don’t know which database server to pick then let me know and I will ask you additional questions to help.
- If a client code that is using your servers is under your control then implement all the fault tolerance logic in there. The simplest way of doing this is the “round robin” - you connect to a random server until you find a server which serves your request. Google “round robin” for more details.
- If a client code is not under your control then you should use BGP load balancing. The main idea of that is that once a server is down another server gets the IP address of the first one, such as on a client side everything is perfect. For more details just google “bgp load balancing”.