feat: batch_mode config option to unlock 3x multiple problem solving speed by reusing openebs and observe namespace (1mins speed up), plus helm pull policy (1mins speed up)

aiopslab/config.yml.example

@@ -13,3 +13,8 @@ qualitative_eval: false
 
 # Flag to enable/disable printing the session
 print_session: false
+
+# Batch mode settings
+batch_mode: false  # Optimizes for running multiple problems sequentially:
+# - Caches Docker images locally to avoid re-pulling
+# - Keeps infrastructure (OpenEBS, Prometheus) running between problems

aiopslab/orchestrator/orchestrator.py

@@ -3,14 +3,14 @@
 
 """Orchestrator class that interfaces with the agent and the environment."""
 
-from aiopslab.service.helm import Helm
 from aiopslab.service.kubectl import KubeCtl
 from aiopslab.session import Session
 from aiopslab.orchestrator.problems.registry import ProblemRegistry
 from aiopslab.orchestrator.parser import ResponseParser
 from aiopslab.utils.status import *
 from aiopslab.utils.critical_section import CriticalSection
 from aiopslab.service.telemetry.prometheus import Prometheus
+from aiopslab.paths import config
 import time
 import inspect
 import asyncio
@@ -203,7 +203,8 @@ async def start_problem(self, max_steps: int):
         # if not self.session.problem.sys_status_after_recovery():
         self.session.problem.app.cleanup()
 
-        if self.session.problem.namespace != "docker":
+        # In batch mode, keep infrastructure running between problems
+        if self.session.problem.namespace != "docker" and not config.get("batch_mode", False):
             self.prometheus.teardown()
             print("Uninstalling OpenEBS...")
             self.kubectl.exec_command("kubectl delete sc openebs-hostpath openebs-device --ignore-not-found")

aiopslab/service/apps/flight_ticket.py

@@ -4,7 +4,6 @@
 from aiopslab.service.helm import Helm
 from aiopslab.service.kubectl import KubeCtl
 from aiopslab.service.apps.base import Application
-from aiopslab.paths import TARGET_MICROSERVICES
 from aiopslab.paths import FLIGHT_TICKET_METADATA
 
 

aiopslab/service/apps/train_ticket.py

@@ -4,7 +4,6 @@
 from aiopslab.service.helm import Helm
 from aiopslab.service.kubectl import KubeCtl
 from aiopslab.service.apps.base import Application
-from aiopslab.paths import TARGET_MICROSERVICES
 from aiopslab.paths import TRAIN_TICKET_METADATA
 
 

aiopslab/service/helm.py

@@ -4,9 +4,9 @@
 """Interface for helm operations"""
 
 import subprocess
-import time
 
 from aiopslab.service.kubectl import KubeCtl
+from aiopslab.paths import config
 
 
 class Helm:
@@ -27,8 +27,13 @@ def install(**args):
         chart_path = args.get("chart_path")
         namespace = args.get("namespace")
         version = args.get("version")
-        extra_args = args.get("extra_args")
+        extra_args = args.get("extra_args", [])
         remote_chart = args.get("remote_chart", False)
+
+        # Apply batch mode if enabled
+        if config.get("batch_mode", False):
+            extra_args = extra_args or []
+            extra_args.append("--set global.imagePullPolicy=IfNotPresent")
 
         if not remote_chart:
             # Install dependencies for chart before installation

aiopslab/service/telemetry/prometheus.py

@@ -4,7 +4,6 @@
 import os
 import json
 import yaml
-import time
 from subprocess import CalledProcessError
 from aiopslab.service.helm import Helm
 from aiopslab.service.kubectl import KubeCtl