Skip to content
PayloadPuffin.
Join the waitlist

A delivery took three tries to reach your destination. Its first two attempts show you why. An attempt is one try at a delivery, usually one signed POST, with a record of what came back.

Attempts are a history, not a status. PayloadPuffin adds one when a try completes and never edits or removes it on its own. Retries and resends add more. The history goes only when its event's retention ends.

What an attempt records

Each attempt in a delivery's attempts list has these fields:

FieldMeaning
idThe attempt ID, starting with att_.
attemptNumber1 for the first attempt, then counting up.
startedAtWhen the attempt started.
completedAtWhen it completed.
statusCodeThe HTTP status the destination returned, or null.
latencyMsHow long the HTTP request took, or null if none was sent.
errorWhy it didn't succeed, or null if it did.
recordsWhether the request and response were kept, one side at a time.

Read the outcome

statusCode and latencyMs together tell you who failed:

statusCodelatencyMsWhat happened
2xxa numberDelivered. The delivery is delivered.
othera numberThe destination answered, but not with a 2xx.
nulla numberThe request was sent, but no response came: a network error or timeout.
nullnullPayloadPuffin failed before sending anything. The destination did nothing wrong.

So a null status code doesn't always mean your server misbehaved. Read latencyMs before you go looking.

Only a 2xx delivers. Most failures are retried, including a 404, a timeout or a transform that throws. Some failures end the delivery at once. Retries has the schedule and the exceptions.

What is not an attempt

PayloadPuffin sometimes waits before it starts an attempt. The destination may have asked it to back off with Retry-After, its concurrency or rate limit may be full, or it may have failed several times in a row. That wait records no attempt and doesn't use up a retry. See Delivery controls for the limits you can set.

What the destination receives

Each attempt is a POST to the destination's URL with the payload as the body. Three headers identify and sign it:

HeaderValue
webhook-idThe delivery ID. The same on every attempt and resend.
webhook-timestampWhen this attempt was signed. Different on every attempt.
webhook-signatureThe signature over the ID, the timestamp, and the exact body.

Because webhook-id stays the same and the timestamp changes, a receiver can recognize a retry it has already processed and still reject an old, replayed signature. Verify signatures shows how to check the signature.

The API and the console show different things

The public API returns each attempt's outcome and its records, but not the request or response bodies. You get attempts inline, in Get a delivery. There is no separate attempts operation.

The console shows the request PayloadPuffin sent and the response it got for each attempt, as a preview with a full download. Credential headers in the sent request are masked to their last four characters. PayloadPuffin keeps the first 1,000,000 bytes of a response body. The delivery succeeds or fails on the status alone.

records.request and records.response tell you whether there's anything to look at. Each has a state:

  • available: written when the attempt finished. If the write never landed after all, opening it says it wasn't found.
  • missing: it should have been kept, but the write failed or there was no room for it.
  • not_recorded: it was never going to be kept. reason says why: disabled when the destination doesn't keep it, not_sent when no request went out, or no_response when no response came back.

Keeping requests and responses is set per destination, and both are on by default. Recording explains the settings.