fix: 503 when private gpt gets ollama service (zylon-ai#2104)

When running private gpt with external ollama API, ollama service returns 503 on startup because ollama service (traefik) might not be ready. - Add healthcheck to ollama service to test for connection to external ollama - private-gpt-ollama service depends on ollama being service_healthy Co-authored-by: Koh Meng Hui <[email protected]>
zhygallo · Oct 17, 2024 · 940bdd4 · 940bdd4
1 parent 5851b02
commit 940bdd4
Showing 1 changed file with 8 additions and 1 deletion.
diff --git a/docker-compose.yaml b/docker-compose.yaml
@@ -29,7 +29,8 @@ services:
       - ollama-cuda
       - ollama-api
     depends_on:
-      - ollama
+      ollama:
+        condition: service_healthy
 
   # Private-GPT service for the local mode
   # This service builds from a local Dockerfile and runs the application in local mode.
@@ -60,6 +61,12 @@ services:
   # This will route requests to the Ollama service based on the profile.
   ollama:
     image: traefik:v2.10
+    healthcheck:
+      test: ["CMD", "sh", "-c", "wget -q --spider http://ollama:11434 || exit 1"]
+      interval: 10s
+      retries: 3
+      start_period: 5s
+      timeout: 5s
     ports:
       - "8080:8080"
     command: