Delivery

The delivery engine is a poll-based worker pool that dequeues pending deliveries, performs HTTP requests, and decides what to do next.

Engine lifecycle

r.Start(ctx)        // start the poll loop and workers
defer r.Stop(ctx)   // wait for in-flight deliveries, then stop

How it works

A ticker fires every PollInterval (default: 1s).
Each tick dequeues up to BatchSize (default: 50) pending deliveries.
Deliveries are dispatched to Concurrency (default: 10) goroutine workers.
Each worker fetches the endpoint and event, performs the HTTP POST, evaluates the result.

HTTP request

Each delivery sends an HTTP POST with these headers:

Header	Value
`Content-Type`	`application/json`
`User-Agent`	`Relay/1.0`
`X-Relay-Event-ID`	Event TypeID
`X-Relay-Event-Type`	Event type name
`X-Relay-Delivery-ID`	Delivery TypeID
`X-Relay-Signature`	`v1=<hex>` HMAC-SHA256
`X-Relay-Timestamp`	Unix timestamp
Custom headers	From endpoint configuration

Decision matrix

After each attempt, the retrier decides the next action:

HTTP Response	Decision
2xx	Delivered -- mark complete
410 Gone	Disable endpoint -- disable endpoint, push to DLQ
400--499 (except 410, 429)	DLQ -- client error won't self-correct
429 Too Many Requests	Retry -- if attempts remain, else DLQ
500--599	Retry -- if attempts remain, else DLQ
0 (network/timeout)	Retry -- if attempts remain, else DLQ

Default retry schedule

var DefaultRetrySchedule = []time.Duration{
    5 * time.Second,   // retry 1
    30 * time.Second,  // retry 2
    2 * time.Minute,   // retry 3
    15 * time.Minute,  // retry 4
    2 * time.Hour,     // retry 5
}

Configuration

Option	Default	Description
`WithConcurrency(n)`	`10`	Worker goroutines
`WithPollInterval(d)`	`1s`	Poll frequency
`WithBatchSize(n)`	`50`	Deliveries per tick
`WithRequestTimeout(d)`	`30s`	HTTP timeout per attempt
`WithMaxRetries(n)`	`5`	Max attempts before DLQ
`WithRetrySchedule(s)`	See above	Backoff intervals