Browse Source

LOEMIND observability: näidis-dashboardid + alertide juhis

ardo 1 month ago
parent
commit
d1babac27b
1 changed files with 33 additions and 0 deletions
  1. 33 0
      dev/LOEMIND_observability.md

+ 33 - 0
dev/LOEMIND_observability.md

@@ -93,6 +93,39 @@ docker exec clickstack mongo --quiet hyperdx --eval \
 # vajadusel: db.sources.updateOne({kind:"metric"},{$set:{"metricTables.exponentialHistogram":"otel_metrics_exponential_histogram"}})
 ```
 
+## Näidis-dashboardid
+Loodud 2026-09-25 skriptiga **`~/scripts/create_hyperdx_dashboards.py`**:
+- **„Ohmu — süsteem (host)"** — RAM vaba, swap vaba, load1, konteinerite mälu (host metrikad)
+- **„Ohmu — mudelid (llama.cpp)"** — genereeritud/prompt-tokenid, aktiivsed päringud, KV-cache
+- **„Ohmu — teenused ja konteinerid"** — `ohmu_service_up`, `ohmu_container_restarts`
+
+Kordamine: `python3 ~/scripts/create_hyperdx_dashboards.py` (`--delete-tests` koristab testid).
+
+**API ligipääs:** HyperDX väline API v2 on **`http://<clickstack-ip>:8000/api/v2`**
+(hostilt konteineri IP kaudu; `clickstack-proxy` ribab `/api` prefiksi, seega `:18080` kaudu ei tööta).
+Access key: `docker exec clickstack mongo --quiet hyperdx --eval 'print(db.users.findOne({email:"admin@odamus.com"}).accessKey)'`.
+
+**Tile-vorming (API):**
+```json
+{"name":"paneel","x":0,"y":0,"w":6,"h":6,"config":{
+  "displayType":"line","sourceId":"<Metrics source id>",
+  "select":[{"valueExpression":"Value","aggFn":"avg","where":"MetricName:\"node_load1\""}],
+  "groupBy":"name"}}
+```
+
+## Alertid
+Alert vajab **teavituskanalit** (`channel`/`channels`) — HyperDX webhook. Ilma kanalita
+API tagastab „Invalid input". Seadistus: UI → Alerts → New Alert, või Webhooks all
+veebikonks (nt n8n → Telegram-relay).
+API: `POST /api/v2/alerts`, kohustuslikud `interval`, `threshold`, `thresholdType`,
+`source` (`inline`|`tile`|`saved_search`) + `channel`.
+
+Näidikud, mida tasub alertida:
+- `ohmu_service_up < 1` (teenus maas) — `thresholdType:"below"`, `threshold:1`, `interval:"5m"`
+- `node_memory_MemAvailable_bytes` liiga madal (RAM otsas)
+- `ohmu_container_restarts` kasvab (konteineri restart-loop)
+- `llamacpp:kv_cache_usage_ratio` > 0.9
+
 ## Seotud
 - Vahejuhtumite audit: `../LOEMIND_seisevad_konteinerid.md`
 - Serveri mudelid: `~/servers/LOEMIND/llama-server.md`