Informational: "Consider adding a PDB"

Comment

Author: @dave (human reviewer) PR: Adds a new Deployment for the notifications service

This Deployment has no PodDisruptionBudget. During a node drain, all replicas could be evicted simultaneously. Worth adding in a follow-up to protect against that.

---

Classification: INFORMATIONAL

Reason: The suggestion is valid but explicitly marked as "follow-up" by the reviewer. Adding a PDB requires knowing the correct minAvailable value, which requires load testing this new service first. Out of scope for an initial rollout PR.

---

Fix: none

---

Reply posted on thread

Agreed — a PDB is the right call once we have a baseline for this service.

Adding it now would require picking a minAvailable value blind, which risks either breaking rolling deploys (too high) or providing no real protection (too low). The plan is to run load tests in staging during the first week, then add the PDB in a follow-up PR with measured values.

Tracked in issue #112.

ℹ️ Thread resolved — no code change needed.

---

When a human would triage this manually

1. Reviewer said "in a follow-up" — this is a suggestion, not a blocker
2. No existing PDB to compare against for this new service
3. Correct response: acknowledge, explain why not now, create a tracking issue, resolve