Per-container Neo4j probing + ServerONLY PDB entry (v0.9.0)

Mirrors the rancher: block pattern from v0.6.0 (caches) and
v0.8.0 (apps) onto Neo4jServiceSpec. Each cluster member gets its
own SHOW DATABASES + count probe directly at
http://<primaryIpAddress>:7474 rather than going through the LB.

YAML:
- PDB Neo4j (pdb.virtualflybrain.org, 1s39, scale 3) gains a
  rancher: block.
- New "PDB Neo4j -- ServerONLY (1s262)" entry (scale 2) for the
  server-only PDB replica.
- KB Neo4j stays LB-fronted (single instance).

Storage: new container column on neo4j_history with automatic
migration (same pattern as cache_history in v0.6.0 and
app_history in v0.8.0). neo4j_series aggregates via
MAX(node_count) across containers per ts.

Page card: cluster summary at the top with ok-count, max node
count and a "mixed" badge when members disagree, plus a per-
container table. Falls back to LB probe transparently when the
Rancher API call fails.

/api/neo4j response reshapes to summary + containers[].

Verified end-to-end: 3 PDB + 2 ServerONLY containers enumerated
and probed in parallel; KB still queries cleanly via the LB
(1,621,245 nodes online); migration of legacy neo4j_history adds
the container column without data loss.

Author: Robbie1977

CHANGELOG.md

@@ -2,6 +2,24 @@
 
 All notable changes to vfb-status are recorded here. The format is loosely based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/) and the project follows [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.9.0] — 2026-06-04
+
+Per-container probing for `neo4j_services` + new ServerONLY PDB entry.
+
+### Added
+
+- Optional `rancher:` block on `neo4j_services` entries (default `container_port: 7474`, `container_scheme: http`). When set, every probe enumerates the service's running containers via the Rancher v1 API and runs the two-stage `SHOW DATABASES` + `MATCH count` check against each container's own HTTP listener at `http://<primaryIpAddress>:7474`. Falls back to the LB-fronted single probe if the API isn't configured or fails.
+- New `container` column on `neo4j_history` (auto-migrated on existing DBs) + matching `idx_neo4j_service_container_ts` index.
+- PDB Neo4j (`pdb.virtualflybrain.org`, service `1s39`, scale 3) now probes each cluster member individually.
+- New `PDB Neo4j — ServerONLY (1s262)` entry (scale 2) — the server-only replica that doesn't run the app sidecar. Surfaces it alongside the main PDB so any divergence between primary and replica is immediately visible.
+- KB Neo4j stays single-LB (no `rancher:` block) — it's a single-instance service so per-container probing adds no information.
+
+### Changed
+
+- Neo4j card layout now mirrors caches/apps: cluster summary at the top (member count, ok count, max node count + `⚠ mixed` flag if containers report different totals, average latency) plus a per-container breakdown table.
+- `neo4j_series` aggregates across containers per ts via `MAX(node_count)` — picks the leader's count when members disagree so a stale follower doesn't drag the chart down.
+- `/api/neo4j` now returns `summary` + `containers[]` per service (existing top-level fields removed — clients reading the old shape need updating).
+
 ## [0.8.2] — 2026-06-01
 
 ### Added
@@ -203,6 +221,7 @@ Initial release. Self-contained Docker uptime tracker for public-facing Virtual
 - Four subdomains (`nas0`, `iip3d`, `nblast`, `abd1-5.catmaid`) ship with `verify_tls: false` because the production cert SAN doesn't cover them. The servers are up; the cert provisioning is a separate problem.
 - Kubernetes nodes are intentionally not handled here — separate checks planned for a later release.
 
+[0.9.0]: https://github.com/VirtualFlyBrain/vfb-status/releases/tag/v0.9.0
 [0.8.2]: https://github.com/VirtualFlyBrain/vfb-status/releases/tag/v0.8.2
 [0.8.1]: https://github.com/VirtualFlyBrain/vfb-status/releases/tag/v0.8.1
 [0.8.0]: https://github.com/VirtualFlyBrain/vfb-status/releases/tag/v0.8.0