chore(autocon5): bump sonda to 1.16.0 + migrate aggregate scrape to /scenarios/metrics

docs/workshop/tour.md

@@ -47,7 +47,7 @@ A single workshop boot registers **76 scenarios** — sonda-server fans each `ki
 | `GET /scenarios/{uuid}` | One scenario's live handle: identity (`id`, `name`), `state`, `elapsed_secs`, plus an embedded `stats` block. |
 | `GET /scenarios/{uuid}/stats` | Per-scenario emission counters: `total_events`, `current_rate`, `target_rate`, `bytes_emitted`, `errors`, `consecutive_failures`, gap/burst state, and `last_successful_write_at`. |
 | `GET /scenarios/{uuid}/metrics` | The single Prometheus-text sample this scenario is emitting right now. Drained on read. |
-| `GET /metrics[?label=k:v]` | Aggregate Prometheus-text snapshot across every running scenario. Optional `label=key:value` filter narrows to one device. Snapshot semantics — multiple consumers (Telegraf and a curl) can read the same bytes concurrently. |
+| `GET /scenarios/metrics[?label=k:v]` | Aggregate Prometheus-text snapshot across every running scenario. Optional `label=key:value` filter narrows to one device. Snapshot semantics — multiple consumers (Telegraf and a curl) can read the same bytes concurrently. |
 | `POST /scenarios` | Register a new scenario. The cascade flap (`nobs autocon5 flap-interface`) is one of these. |
 | `DELETE /scenarios/{uuid}` | Stop and unregister a scenario. |
 
@@ -99,10 +99,10 @@ curl -s http://localhost:8085/scenarios | jq '.scenarios[0]'
 }
 ```
 
-To see the actual Prometheus-text samples Telegraf reads — the byte-for-byte input, before any rename or normalization — curl the aggregate `/metrics` endpoint and filter to one device with `label=key:value`. The endpoint is snapshot-style: Telegraf scrapes it every 10 seconds and you can read it concurrently without stealing samples from anyone.
+To see the actual Prometheus-text samples Telegraf reads — the byte-for-byte input, before any rename or normalization — curl the aggregate `/scenarios/metrics` endpoint and filter to one device with `label=key:value`. The endpoint is snapshot-style: Telegraf scrapes it every 10 seconds and you can read it concurrently without stealing samples from anyone.
 
 ```bash
-curl -s 'http://localhost:8085/metrics?label=source:srl1' | grep '^srl_bgp_oper_state{' | grep '10.1.99.2' | head -1
+curl -s 'http://localhost:8085/scenarios/metrics?label=source:srl1' | grep '^srl_bgp_oper_state{' | grep '10.1.99.2' | head -1
 ```
 
 ```text
@@ -114,7 +114,7 @@ Note the metric name (`srl_bgp_oper_state`) and the `source="srl1"` tag — that
 srl2 uses the SNMP shape, where the device tag is `agent_host`:
 
 ```bash
-curl -s 'http://localhost:8085/metrics?label=agent_host:srl2' | grep '^bgpPeerState{' | grep '10.1.2.1' | head -1
+curl -s 'http://localhost:8085/scenarios/metrics?label=agent_host:srl2' | grep '^bgpPeerState{' | grep '10.1.2.1' | head -1
 ```
 
 ```text
@@ -145,7 +145,7 @@ Drop `--kind composable` to also list runnable scenarios in the catalog, or swap
 
 ### Where you'll see this in the workshop
 
-- **Part 1** — When you query Prometheus and see `bgp_oper_state{device="srl1"}` return three rows with one stuck at `5` instead of `1`, that's a value Sonda is generating right now. `curl 'http://localhost:8085/metrics?label=source:srl1'` shows you the raw side of that same number.
+- **Part 1** — When you query Prometheus and see `bgp_oper_state{device="srl1"}` return three rows with one stuck at `5` instead of `1`, that's a value Sonda is generating right now. `curl 'http://localhost:8085/scenarios/metrics?label=source:srl1'` shows you the raw side of that same number.
 - **Part 3** — `nobs autocon5 flap-interface` and `nobs autocon5 incident` both POST cascade scenarios to this server.
 - **Advanced** — When you write your own scenario, you'll POST it here.
 

workshops/autocon5/.env.example

@@ -8,7 +8,7 @@
 # Image pins (optional). Leave commented to use the defaults baked into the
 # compose file.
 # ---------------------------------------------------------------------------
-# SONDA_IMAGE=ghcr.io/davidban77/sonda:1.3.0
+# SONDA_IMAGE=ghcr.io/davidban77/sonda:1.16.0
 # PROMETHEUS_IMAGE=docker.io/prom/prometheus:v2.52.0
 # LOKI_IMAGE=docker.io/grafana/loki:3.1.1
 # GRAFANA_IMAGE=docker.io/grafana/grafana:13.0.1

workshops/autocon5/docker-compose.yml

@@ -34,11 +34,11 @@ services:
   # =========================================================================
 
   # Both devices register their baseline scenarios on sonda-server. Each
-  # telegraf scrapes the aggregate /metrics endpoint, filtered by a device
+  # telegraf scrapes the aggregate /scenarios/metrics endpoint, filtered by a device
   # label (`source` for srl1's gNMI shape, `agent_host` for srl2's SNMP shape).
   sonda-server:
     container_name: sonda-server
-    image: ${SONDA_IMAGE:-ghcr.io/davidban77/sonda:1.14.0}
+    image: ${SONDA_IMAGE:-ghcr.io/davidban77/sonda:1.16.0}
     command: ["--port", "8080", "--bind", "0.0.0.0", "--catalog", "/catalog"]
     volumes:
       - ./sonda/catalog:/catalog:ro
@@ -68,7 +68,7 @@ services:
   # to Loki; srl2 streams emit RFC 5424 syslog UDP to Vector (vector:1514).
   sonda-logs:
     container_name: sonda-logs
-    image: ${SONDA_IMAGE:-ghcr.io/davidban77/sonda:1.14.0}
+    image: ${SONDA_IMAGE:-ghcr.io/davidban77/sonda:1.16.0}
     entrypoint: ["/sonda"]
     command: ["run", "/catalog/all-logs.yaml"]
     volumes:
@@ -87,7 +87,7 @@ services:
   # canonical `interface_*` / `bgp_*` series in Prometheus.
   # =========================================================================
 
-  # telegraf-srl1: scrapes the aggregate /metrics endpoint filtered by
+  # telegraf-srl1: scrapes the aggregate /scenarios/metrics endpoint filtered by
   # `source:srl1` (gNMI-raw `srl_*`), renames to canonical (`interface_*`,
   # `bgp_*`), exposes a Prom scrape endpoint on :9005.
   telegraf-srl1:
@@ -111,7 +111,7 @@ services:
       sonda-setup:
         condition: service_completed_successfully
 
-  # telegraf-srl2: scrapes the aggregate /metrics endpoint filtered by
+  # telegraf-srl2: scrapes the aggregate /scenarios/metrics endpoint filtered by
   # `agent_host:srl2` (SNMP-raw `ifHC*`, `bgpPeer*`), renames to canonical,
   # exposes a Prom scrape endpoint on :9005.
   telegraf-srl2:

workshops/autocon5/docs/data-pipelines.md

@@ -213,7 +213,7 @@ Workshop CLI (host machine)                       Prefect flows (in-container)
               │  http://localhost:8085)                         │  http://sonda-server:8080)
               ▼                                                  ▼
               ┌──────────────────────────────────────────────────┐
-              │ sonda-server v1.3.0                              │
+              │ sonda-server v1.16.0                             │
               │  POST /events handler                            │
               │   • validates payload                            │
               │   • forwards to sink config inline               │
@@ -318,7 +318,7 @@ docker compose --project-name autocon5 restart sonda-setup
 
 ## Why `incident.py` uses remote_write
 
-Every other workshop metric flows through the pull path — **sonda emits → Telegraf scrapes `/metrics` → Prometheus scrapes Telegraf**. `nobs autocon5 incident` is the exception: its cascade body uses `encoder/sink: remote_write` and pushes directly to Prometheus, bypassing Telegraf. This is intentional, not legacy.
+Every other workshop metric flows through the pull path — **sonda emits → Telegraf scrapes `/scenarios/metrics` → Prometheus scrapes Telegraf**. `nobs autocon5 incident` is the exception: its cascade body uses `encoder/sink: remote_write` and pushes directly to Prometheus, bypassing Telegraf. This is intentional, not legacy.
 
 ### Why the exception exists
 

workshops/autocon5/guides/part-1-telemetry-and-queries.md

@@ -40,15 +40,15 @@ The two devices speak different protocols at the source:
 - **`srl1` emits gNMI** — that's the telemetry shape SR Linux puts on the wire natively. Field names like `srl_interface_oper_state`, tags like `source`. **Telegraf-srl1** scrapes this raw shape, renames `srl_*` to canonical (`interface_*`, `bgp_*`) and `source` to `device`. Out the other side: the shared schema this workshop's dashboards and alerts speak.
 - **`srl2` emits SNMP** — the classic shape from IF-MIB / BGP4-MIB. Field names like `ifOperStatus`, `ifHCInOctets`; tags like `agent_host`, `ifDescr`. **Telegraf-srl2** scrapes the raw SNMP shape and renames every field and every tag to the same canonical schema. Out the other side: byte-for-byte identical to what srl1 produces.
 
-Both raw shapes live on `sonda-server` (the lab's synthetic-telemetry runtime). Each Telegraf scrapes its device's per-scenario `/metrics` endpoints on a 10-second cadence — same scrape pattern Prometheus would use against real exporters in production.
+Both raw shapes live on `sonda-server` (the lab's synthetic-telemetry runtime). Each Telegraf scrapes its device's per-scenario `/scenarios/metrics` endpoints on a 10-second cadence — same scrape pattern Prometheus would use against real exporters in production.
 
 #### See the raw shape, before Telegraf touches it
 
 > Your senior pulls up a terminal. *"You don't have to take the bullet list on faith. Each layer of the pipeline has its own URL — just click them and see the same fact in three different shapes."*
 
 The pipeline has three layers you can inspect directly from your browser:
 
-1. **Raw gNMI from srl1** (sonda-server, before Telegraf): <http://localhost:8085/metrics?label=source:srl1>
+1. **Raw gNMI from srl1** (sonda-server, before Telegraf): <http://localhost:8085/scenarios/metrics?label=source:srl1>
 
     Look for `srl_*` metric names and the `source="srl1"` tag. This is what an SR Linux device emits on its gNMI stream. For example:
 
@@ -58,7 +58,7 @@ The pipeline has three layers you can inspect directly from your browser:
 
     The pack `workshops/autocon5/sonda/catalog/srlinux-gnmi-interfaces-raw.yaml` lists every metric in this shape.
 
-2. **Raw SNMP from srl2** (sonda-server, before Telegraf): <http://localhost:8085/metrics?label=agent_host:srl2>
+2. **Raw SNMP from srl2** (sonda-server, before Telegraf): <http://localhost:8085/scenarios/metrics?label=agent_host:srl2>
 
     Different shape entirely — IF-MIB names (`ifOperStatus`, `ifHCInOctets`) and the `agent_host="srl2"` tag. For example:
 
@@ -80,9 +80,9 @@ The pipeline has three layers you can inspect directly from your browser:
 
     A single PromQL query for `interface_oper_state` returns rows from both devices in the same shape. The vendor difference is invisible at this layer.
 
-??? info "Why the sonda `/metrics` endpoint is safe for two readers at once"
+??? info "Why the sonda `/scenarios/metrics` endpoint is safe for two readers at once"
 
-    `sonda-server` exposes two shapes of metric endpoint: the **aggregate** `/metrics?label=key:value` you just used, and a **per-scenario** `/scenarios/{id}/metrics` for a single scenario by ID.
+    `sonda-server` exposes two shapes of metric endpoint: the **aggregate** `/scenarios/metrics?label=key:value` you just used, and a **per-scenario** `/scenarios/{id}/metrics` for a single scenario by ID.
 
     - The aggregate endpoint is **snapshot-style**: each scrape gets a consistent picture without consuming anything. Telegraf reads it every 10 seconds; you can read it concurrently from your browser; both see the same bytes.
     - The per-scenario endpoint is **drain-on-read**: each read consumes the scenario's emission buffer. Telegraf doesn't use this endpoint precisely because two consumers can't share a drain-on-read buffer without racing.
@@ -646,7 +646,7 @@ UPDOWN log lines start appearing in the live stream within seconds of the first
 
         Click each URL and grep for one specific metric:
 
-        - <http://localhost:8085/metrics?label=agent_host:srl2> — raw SNMP (pre-Telegraf): `bgpPeerState`, `ifHCInOctets`, `agent_host=srl2`
+        - <http://localhost:8085/scenarios/metrics?label=agent_host:srl2> — raw SNMP (pre-Telegraf): `bgpPeerState`, `ifHCInOctets`, `agent_host=srl2`
         - <http://localhost:9006/metrics> — telegraf-srl2's normalized output: `bgp_oper_state`, `interface_in_octets`, `device=srl2`
         - <http://localhost:9090/graph?g0.expr=bgp_oper_state%7Bdevice%3D%22srl2%22%7D&g0.tab=1> — Prometheus stores the same data after one more scrape hop
 

workshops/autocon5/guides/part-2-dashboards.md

@@ -232,8 +232,8 @@ Worth noting: `srl1` and `srl2` arrive through different upstream pipelines (gNM
 
     **See it yourself — five URLs walk the three layers of each pipeline:**
 
-    1. **Raw gNMI from srl1** (sonda-server, before Telegraf): <http://localhost:8085/metrics?label=source:srl1>. Look for `srl_*` metric names (`srl_interface_oper_state`, `srl_bgp_oper_state`) and the `source="srl1"` tag — what an SR Linux device emits on its gNMI stream.
-    2. **Raw SNMP from srl2** (sonda-server, before Telegraf): <http://localhost:8085/metrics?label=agent_host:srl2>. Look for the IF-MIB / BGP4-MIB names (`ifHCInOctets`, `bgpPeerState`, `cbgpPeerOperStatus`) and the `agent_host="srl2"` tag — the classic SNMP shape.
+    1. **Raw gNMI from srl1** (sonda-server, before Telegraf): <http://localhost:8085/scenarios/metrics?label=source:srl1>. Look for `srl_*` metric names (`srl_interface_oper_state`, `srl_bgp_oper_state`) and the `source="srl1"` tag — what an SR Linux device emits on its gNMI stream.
+    2. **Raw SNMP from srl2** (sonda-server, before Telegraf): <http://localhost:8085/scenarios/metrics?label=agent_host:srl2>. Look for the IF-MIB / BGP4-MIB names (`ifHCInOctets`, `bgpPeerState`, `cbgpPeerOperStatus`) and the `agent_host="srl2"` tag — the classic SNMP shape.
     3. **Telegraf-srl1's normalized output**: <http://localhost:9005/metrics>. The `srl_*` names are now plain `interface_*` / `bgp_*`, and the `source` tag has been renamed to `device`. Same data, canonical shape.
     4. **Telegraf-srl2's normalized output**: <http://localhost:9006/metrics>. The SNMP names (`ifHCInOctets`, etc.) are now also `interface_*` / `bgp_*`, and `agent_host` is now `device`. Identical to telegraf-srl1's output above — except for one label we keep on purpose: `collection_type=gnmi` vs `collection_type=snmp`, so you can debug which pipeline a sample came from.
     5. **Final view in Prometheus**: <http://localhost:9090/graph?g0.expr=interface_oper_state%7Bdevice%3D~%22srl1%7Csrl2%22%7D&g0.tab=1>. A single query for `interface_oper_state{device=~"srl1|srl2"}` returns rows from both devices in the same shape — the vendor difference is invisible at this layer.

workshops/autocon5/sonda/catalog/srl1-metrics.yaml

@@ -1,5 +1,5 @@
 # srl1 telemetry — RAW gNMI shape (`srl_*` names + `source` tag).
-# Registered on sonda-server; telegraf-srl1 scrapes the aggregate /metrics
+# Registered on sonda-server; telegraf-srl1 scrapes the aggregate /scenarios/metrics
 # endpoint and normalizes the shapes to the canonical schema.
 # Broken-on-purpose: BGP peer 10.1.99.2 and interface ethernet-1/11.
 #

workshops/autocon5/sonda/scripts/sonda-setup.sh

@@ -3,7 +3,7 @@
 #
 # Runs as an init container. Waits for sonda-server, POSTs each `*.yaml` in
 # `$SCENARIOS_DIR` (server resolves any `pack:` refs via its own `--catalog`),
-# and exits. Telegraf scrapes the aggregate /metrics endpoint filtered by
+# and exits. Telegraf scrapes the aggregate /scenarios/metrics endpoint filtered by
 # device label, so no per-device ID file is needed.
 
 set -e

workshops/autocon5/telegraf/telegraf-srl1.conf.toml

@@ -5,14 +5,14 @@
 ###############################################################################
 #                            DATA COLLECTION                                  #
 ###############################################################################
-# Scrape sonda's aggregate /metrics, filtered to srl1 (gNMI shape).
+# Scrape sonda's aggregate /scenarios/metrics, filtered to srl1 (gNMI shape).
 # `include_state=running,unresolved,held` is the inversion-pattern filter:
 # baselines sit in `unresolved` (with `if_unresolved: open`) at rest, the
 # cascade signal flips them to `paused` while a cascade is running, and
 # `held` admits snap_to-equipped scenarios that have frozen at a recovery
 # value (Pattern C, sonda PR #440 / 1.13.2+).
 [[inputs.prometheus]]
-  urls = ["http://sonda-server:8080/metrics?label=source:srl1&include_state=running,unresolved,held"]
+  urls = ["http://sonda-server:8080/scenarios/metrics?label=source:srl1&include_state=running,unresolved,held"]
   metric_version = 2
   tagexclude = ["url", "host"]
 

workshops/autocon5/telegraf/telegraf-srl2.conf.toml

@@ -5,14 +5,14 @@
 ###############################################################################
 #                            DATA COLLECTION                                  #
 ###############################################################################
-# Scrape sonda's aggregate /metrics, filtered to srl2 (SNMP-raw shape).
+# Scrape sonda's aggregate /scenarios/metrics, filtered to srl2 (SNMP-raw shape).
 # `include_state=running,unresolved,held` is the inversion-pattern filter:
 # baselines sit in `unresolved` (with `if_unresolved: open`) at rest, the
 # cascade signal flips them to `paused` while a cascade is running, and
 # `held` admits snap_to-equipped scenarios that have frozen at a recovery
 # value (Pattern C, sonda PR #440 / 1.13.2+).
 [[inputs.prometheus]]
-  urls = ["http://sonda-server:8080/metrics?label=agent_host:srl2&include_state=running,unresolved,held"]
+  urls = ["http://sonda-server:8080/scenarios/metrics?label=agent_host:srl2&include_state=running,unresolved,held"]
   metric_version = 2
   tagexclude = ["url", "host"]