OpenSpec: разнести openspec по проектам

2026-09-29 09:55:09 +00:00 · 2026-09-11 17:31:01 +00:00
parent 9f2619039a
commit 3c5e9e5a71
25 changed files with 1893 additions and 0 deletions
@@ -0,0 +1,84 @@
+# vinograd-wan-monitoring Specification
+
+## Purpose
+TBD - created by archiving change vinograd-rostelecom-channel-monitoring. Update Purpose after archive.
+
+## Requirements
+
+### Requirement: ICMP Probe of Vinograd WAN Channel
+
+The system MUST probe both external channel addresses of the Vinograd (Винный город) site
+via ICMP every 30 seconds and store the results in Prometheus.
+
+| Address | Role |
+|---|---|
+| 83.239.50.145 | Gateway (шлюз Ростелеком) |
+| 83.239.50.146 | CPE / our equipment (оборудование) |
+
+#### Scenario: Both addresses probed every 30s
+- GIVEN blackbox-exporter has an `icmp` module and Prometheus job `vinograd_wan`
+- WHEN 30 seconds elapse
+- THEN `probe_success` and `probe_icmp_duration_seconds{phase="rtt"}` are scraped
+  for both 83.239.50.145 and 83.239.50.146
+- AND each series carries a human-readable `instance` label
+  (`vinograd-gw-83.239.50.145`, `vinograd-cpe-83.239.50.146`)
+
+#### Scenario: Probe failure
+- GIVEN an address does not answer ICMP (e.g. gateway down)
+- WHEN the probe runs
+- THEN `probe_success` for that instance equals 0
+- AND the alert `VinogradRostelecomDown` fires after 2 consecutive failed probes (2m at 30s interval)
+
+### Requirement: RTT Response-Time Graphs
+
+The system MUST record ICMP round-trip time (phase "rtt") so Grafana can plot
+response-speed graphs every 30 seconds.
+
+#### Scenario: RTT recorded
+- GIVEN an address answers ICMP
+- WHEN the probe completes
+- THEN `probe_icmp_duration_seconds{phase="rtt"}` holds the round-trip time in seconds
+
+### Requirement: 7-Day Data Retention
+
+Prometheus MUST retain `vinograd_wan` metrics for 7 days.
+
+#### Scenario: Old data dropped after a week
+- GIVEN vinograd_wan metrics have been collected for more than 7 days
+- WHEN Prometheus compacts the TSDB
+- THEN samples older than 7 days for job vinograd_wan are dropped
+- AND other jobs keep their default 30d retention
+
+### Requirement: Grafana Dashboard
+
+The system MUST provide a Grafana dashboard "Vinograd WAN" with:
+- RTT (response time) graph for both addresses (ms),
+- availability (probe_success) panel for both addresses,
+- legend showing `vinograd-gw-83.239.50.145` / `vinograd-cpe-83.239.50.146`.
+
+#### Scenario: Dashboard shows data
+- GIVEN Grafana has the Vinograd WAN dashboard provisioned
+- WHEN a user opens it
+- THEN it shows the RTT graph and availability of both channel addresses
+
+### Requirement: Vinograd WAN dashboard opens without datasource errors
+
+The Vinograd WAN dashboard (`/d/vinograd-wan/vinograd-wan`) MUST open and render
+all panels WITHOUT the error "Datasource __grafana__ was not found".
+
+- The dashboard JSON MUST NOT reference the built-in `__grafana__` datasource in
+  its `annotations.list` (it is not registered in this Grafana's database).
+- `annotations.list` MUST be empty (`[]`), matching the working
+  `garage-cluster.json` dashboard.
+
+#### Scenario: Dashboard renders without datasource error
+
+- **WHEN** a user opens `https://grafana.nixg.ru/d/vinograd-wan/vinograd-wan`
+- **THEN** the dashboard loads without the error "Datasource __grafana__ was not found"
+- **AND** all panels render metric data from Prometheus (`uid: Prometheus`)
+
+#### Scenario: Dashboard file stores no __grafana__ reference
+
+- **WHEN** the file `grafana/dashboards/vinograd-wan.json` is parsed
+- **THEN** `annotations.list` is `[]` OR contains no item whose
+  `datasource.uid` equals `__grafana__`