add-eb-global-support

[jgilbert01]

No description provided.

[jgilbert01]

• jumping thru a hoop here with this condition to keep unit tests backwards compatible
• otherwise it will be necessary to assert EndpointId: undefined in many tests

[petermyers]

If we're in the middle of a failover and the publishing is failing, is that considered a retryable error? So in theory it would bypass faulting and retry until the failover is complete and we can continue publishing to the global EB?
I think right now its checking based on the smithy error classifications and we use:
(isThrottlingError(err) || isTransientError(err) || isServerError(err))

[jgilbert01]

If we're in the middle of a failover and the publishing is failing, is that considered a retryable error? So in theory it would bypass faulting and retry until the failover is complete and we can continue publishing to the global EB? I think right now its checking based on the smithy error classifications and we use: (isThrottlingError(err) || isTransientError(err) || isServerError(err))

• right
• global endpoint is powered by r53 primary/secondary failover
• our retry feature is not enabled by default for backwards compatibility
• when enabled the lambda will retry retriable errors, such as 5xx, instead of publishing faults
• the regional failover will usually take about 5 minutes to avoid premature failover
• then the events will flow to the secondary region