We have experienced loss of ingress log events already in a few scenarios, which is an artifact of our current architecture lacking a reliable message archiving tier, i.e. source messages are lost once consumed by a downstream tier.
Most message brokering solutions elsewhere are using the obvious approach to also archive every event to persistent storage before consuming it by (potential several) downstream tiers, usually S3.
This would enable both, recovery from outages without data loss as well as (selective) replay capabilities for testing etc.
This is a generally applicable question, but has been triggered again by https://github.com/cityindex/logsearch/issues/269#issuecomment-30518105 specifically.
We have experienced loss of ingress log events already in a few scenarios, which is an artifact of our current architecture lacking a reliable message archiving tier, i.e. source messages are lost once consumed by a downstream tier.
Most message brokering solutions elsewhere are using the obvious approach to also archive every event to persistent storage before consuming it by (potential several) downstream tiers, usually S3.
This would enable both, recovery from outages without data loss as well as (selective) replay capabilities for testing etc.