docs/ops: fix Prometheus metrics scrape ports

[keoyle52]

Summary

Fix Prometheus scrape target ports in deployments/monitoring/config-prometheus/prometheus.yml to match Quake localdev metrics port offsets.

Issue

Related to #51 (addresses only the Prometheus scrape-port portion — issue #51 bundles several other unrelated problems that remain unfixed).

Change

Corrected execution scrape target ports (9001, 9101, 9201, 9301, 9401) and consensus scrape target ports (29000, 29100, 29200, 29300, 29400) in deployments/monitoring/config-prometheus/prometheus.yml to align with the port allocation logic in crates/quake/src/node.rs (RETH_METRICS_BASE_PORT 9001 + node_index * 100; APP_METRICS_BASE_PORT 29000 + node_index * 100).

Verification

• make lint: PASS
• make build: PASS
• make test-unit: PASS (1616/1619 unit tests passed)
• Runtime Verification: Validated by re-deriving the port formula from crates/quake/src/node.rs, not by an actual running Prometheus instance (Docker runtime not available in local session environment).

Known pre-existing issues (unrelated to this change)

3 arc-test-integration multi-node tests fail locally with timeouts, believed environment-related (see brief §3); unrelated to this change; not cross-checked against upstream CI.

[osr21]

Verified this against main (90d71dc). The execution port corrections are right, but the consensus changes are based on a wrong stride and would break consensus scraping. Details below.

Execution — correct ✅

ExecutionContainer::new_local computes port_offset = node_index * 100 (crates/quake/src/node.rs:344) and sets metrics_port = RETH_METRICS_BASE_PORT + port_offset (9001 base). So 9001 / 9101 / 9201 / 9301 / 9401 is right, and the diff matches.

Consensus — incorrect ❌

Consensus does not use a * 100 stride. ConsensusContainer::new_local(node, subnet_indexes, node_index, port_offset) takes port_offset from its caller, and the caller passes index, not index * 100:

// crates/quake/src/node.rs:442
consensus: ConsensusContainer::new_local(name, subnet_indexes, index, index),

So metrics_port = APP_METRICS_BASE_PORT + port_offset = 29000 + index (node.rs:248), i.e. 29000 / 29001 / 29002 / 29003 / 29004.

That's confirmed downstream by how the ports are actually published to the host — the compose template maps the host port straight onto the fixed container port:

# crates/quake/templates/local/compose.yaml.hbs:139
- "{{consensus.metrics_port}}:29000"   # host = consensus.metrics_port = 29000 + index
# :62
- "{{execution.metrics_port}}:9001"    # host = execution.metrics_port = 9001 + index*100

Prometheus scrapes those same host ports (crates/quake/templates/prometheus.yml.hbs renders host.docker.internal:{{consensus.metrics_port}}).

Net: the consensus targets were already correct at 29000 / 29001 / 29002 / 29003 / 29004. Changing them to 29100 / 29200 / 29300 / 29400 points Prometheus at ports nothing is published on, so nodes 1–4 consensus metrics would silently stop being scraped. Recommend dropping the four consensus edits and keeping only the execution ones.

Side note (optional)

deployments/monitoring/config-prometheus/prometheus.yml is a hand-maintained duplicate of the generator at crates/quake/templates/prometheus.yml.hbs, which derives both port sets from metrics_port and can't drift. If the intent is a stable localdev config, it might be worth generating this file from the template (or referencing it) rather than editing the static copy, to avoid exactly this kind of divergence.