Skip to content

2016

When core data changes, something else has to happen. A cache entry must go, a search index must be updated, and an audit log wants to know what the value was before. I had one mechanism for all three in my previous job, and missed it in the new one. So I built it again from scratch.

Waiting for transaction propagation

In the land of traditional ACID DBs, I often encounter code that is trying to collaborate with some external system using a queue or notification system. The problem is that it ends up sending the notification to the external system before the data has been committed to the central database. Then when the external system tries to query for data based on the notification, the data is not there yet (a different or outside a transaction).