Generic, simple retry module with exponential backoff.
- Easy to understand the different states
- Safe functions by design (see below)
- Exponential backoff
- Supports timeouts
- Fatal errors
- Retries
npm install try-again
var Again = require('try-again')
var again = Again({
retries: 8,
max: 10000,
jitter: .2,
factor: 2,
min: 100
})
// this function will get re-called each time there is
// failure, unless retries is 0 or fatal(...) is called
var client = again(function (success, failure, fatal) {
var client = new Client(url)
client.once('connected', success)
client.once('close', failure)
client.once('error', failure)
return client
}, status, failed)
// this function will get called whenever one of the
// 3 functions: success, failure, or fatal get called.
// This function is often used to update connection state.
function status (err) {
if (err) {
// there was a failure
// update connection state accordingly
} else {
// there was a success
// update connection state accordingly
}
}
// this function is used when the retries have been
// exhausted or the fatal function has been called.
// at this point, there will be no more retries and
// you should consider crashing the process.
function failed (err) {
console.error({
message: 'aborting, tried too many times'
error: err.stack || err
})
}
You don't need to know the details of how this works to use this module, but if you're interested in knowing how the different states may interact, this should help explain things.
Everything inside the again
function should be idempotent
The function inside again
will be called multiple times when there is a failure, so it's important that you don't have existing event emitters and other things hanging around. You should create a new client inside this function each time.
The success
function only works once and only if failure
has not already been called
failure
may be called after success
has been called, but success
will be a noop if failure
has been called. This is to prevent multiple success
functions from running if the connection is eventually successful.
The status
function will be called each time there is an update, either a successful connection or a failure
If there is a failure, the err
parameter will be populated. This function may be called multiple times. It's a good place for logging connection status and setting "connected" state.
The failed
function will only be called if the number of attempts to connect have exceeded the retries option
If the failed
function is called, it won't try anymore. You probably want to handle this case by failing fast and killing the process.
The fatal
function may be used to trigger the failed
function even if there are retries available.
You can use the fatal function to skip retrying. Like the failure
function, the fatal
function may be called after an initial success
function, but cannot be called after the failure
function has previously been called for that cycle.
npm install
make test
MIT