Advanced
Open
Pro
Blackboard vs. Isolated Workers for a Multi-Service Incident Investigation
During a production incident, you want to spin up one agent per affected service (5 services) to independently investigate logs and metrics, looking for the root cause. A colleague proposes giving all five agents read/write access to a shared "incident scratchpad" document so they can see what each other has found in real time, arguing it will help them converge on the root cause faster.
- What does the shared scratchpad buy you here that fully isolated workers reporting back to a supervisor would not?
- What does it cost, concretely?
- Propose a design, and justify whether it uses a scratchpad, fully isolated workers, or something else.
Share this question