This section collects brief definitions of some of the technical terms used in the documentation for Convoy.
Projects are used to create logical contexts or separate environments (dev, staging & production). Different projects can also be created for different teams each with its own login details on the same convoy deployment.
An endpoint represents a target URL to receive events. An endpoint can be in either of these states -
pending. When an endpoint is in the
inactive state all events sent will be saved but not dispatched until the endpoint is re-enabled. These events are known set to the
An event represents a specific event triggered by your system. Convoy persists events sent to dead endpoints with a status -
Discarded. This enables users re-activate their endpoints and easily retry events without the need to re-trigger the events from your systems.
An event delivery is resource that tracks successive attempts to deliver the event payload to each endpoint. Multiple event deliveries can be created for a single event, this is influenced by the number of subscriptions that are matched to it.
Events are sent to an endpoint depending on the event type, which is defined when creating the endpoint defaulting to
"*" if not set, which is a catch all for all events. An endpoint can define multiple event types, as such it will receive an event from all those events. Event types are matched using direct string comparison and are case sensitive. Support for regex event matching is planned.
A delivery attempt represents a single attempt to dispatch an event to an endpoint. Specifically, it contains two things - Request Headers & Payload, Response Headers & Payload. Convoy records this information for every retry attempt sent. The UI currently shows only the last delivery attempt. The number of delivery attempts and retry strategy can be configured per group.
A dead endpoint is an endpoint that failed consecutively to acknowledge events. Currently, we define consecutively failures as at least one event as maxed out it's retry limit to the maximum configured. In the future, we should support different consecutive failure strategies.
Secrets are used to sign the payload when sending events to an endpoint. If you don't supply a secret convoy will generate one for you.
We have found out that most implementations use -
SHA512. However, convoy also supports the following hash functions:
A replay attack occurs when an attacker intercepts a valid network payload with the intent of fraudulently re-transmitting the payload. Convoy supports replay attack prevention by including a timestamp in the request header under the key
Convoy-Timestamp. This timestamp is also included in the signature-header and is signed together with the request body using the endpoint secret. Therefore, an attacker cannot change the timestamp without invalidating the signature. Take the following steps to verify your signature and prevent replay attacks;
- Extract the timestamp and the signed signature-header from the request header, extract the request body.
- Prepare a string by concatenating the timestamp followed by a
,and the request body.
- Generate a signature of the concatenated string using the endpoint secret and your hashing algorithm (e.g
- Compare the newly generated signature with the value in the signature-header, if the signatures match, check the time interval between the timestamp and the current time. In your system, set a tolerance on this time interval to prevent replay attacks.
Rate Limiting Endpoints
While you are guaranteed you'll be able to receive events as fast as possible using convoy, your customers might not be able to handle events coming to their systems at the same rate which might cause a disruption of service on their end. You can control the number events you want to send to an endpoint by setting a rate limit and a rate limit duration on each endpoint. The default is
1m i.e. 5,000 requests per minute.
When an endpoint is experiencing temporary disruption of service, events sent to them might fail to require you to retry them. Convoy allow you to set the number of attempts to a particular endpoint and how to initiate the retry. Convoy supports two retry strategies
default: retries are done in linear time. It's best to set a reasonable number of attempts if the duration is short.
exponential-backoff: retries events while progressively increasing the time before the next attempt. The default schedule looks like this:
- 10 seconds
- 30 seconds
- 1 minute
- 3 minutes
- 5 minutes
- 10 minutes
- 15 minutes Retry strategies are configured per group