1 comment

[ 3.3 ms ] story [ 9.2 ms ] thread
Author here!

Sharing a technical write-up of our recent incident where our service repeatedly crashed due to a poison pill event in the async workers.

Interesting from a general incident coordination angle, but also in terms of mitigations some of which apply generally to web apps, such as splitting work by category/type for improved reliability.

Hope you enjoy, feedback always welcome!