For the event types we support (see the Events Reference), we guarantee at least once delivery of all events. That means we consider it an outage if an event is not delivered, but the receiver (you) are expected to handle it gracefully if we deliver it more than once.
We do not guarantee that events are delivered in the same order that the changes happened in Benchling. In practice, they generally are, but anything listening to them should account for this possibility.
Most events are delivered within a few seconds of the change happening in Benchling that triggers them.
If the system is under a high load, or if the event is triggered as part of a large bulk transaction (e.g. bulk registering entities), event delivery latency may be significantly higher. For example, it may take on the order of 20 seconds to emit 1000 events corresponding to a bulk transaction, and on the order of 1 minute to emit 10 thousand.
EventBridge subscriptions have 3 statuses:
Active subscription requires no further action; its setup has been completed in AWS.
Pending subscription has been created by Benchling as a partner event source in AWS, but it lacks an associated event-bus in AWS. This setup must be completed within 12 days, steps are in the Getting Started Guide.
Expired subscription was created at least 12 days ago, but no event-bus was associated. Expired event subscriptions will not be present as a partner event source in AWS. Click resubscribe in the events settings page (my-tenant.benchling.com/event-subscriptions) to re-create the partner event source and return the subscription to the
Events are able to be retrieved from the List Events API for up to 2 weeks after they have been issued.
We currently do not have plans to support any additional delivery methods, but please contact Benchling support if you have a specific interface or use case that's needed!
In the event of an outage, Benchling will notify customers of the affected time period. Events are able to be retrieved from the List Events API for up to 2 weeks after they have been issued and can be filtered by time. Steps to recover would be:
- Receive outage window from Benchling
- Query the dropped events from the List Events API
- Route the events to the integration(s) in your infrastructure
We do not resend events over EventBridge. Any filtering/routing of these retrieved events will have to be handled either manually or through your infrastructure.
Updated about 1 year ago