SidhaGarg/Cloud-DevOps-RLEnv

1# Cloud DevOps RLEnv2 3Use this page as the operator playbook in the /web interface.4 5## Core Loop6 71. Triage quickly.82. Build evidence from logs and metadata.93. Apply the minimum safe remediation.104. Verify resolution and stop.11 12Why this matters: each action has a cost (`-0.01`), so shorter correct trajectories score higher.13 14## Command Cheat Sheet15 16- `list_resources`17: Start here to map real targets vs decoys.18 19- `describe_resource(resource_id)`20: Inspect config/state details.21 22- `view_logs(resource_id)`23: Primary source of root-cause evidence.24 25- `query_metadata(parameters={"ip_address": "..."})`26: Resolve IP-only clues to resource IDs (mandatory multi-hop step in medium/hard).27 28- `update_security_group(resource_id, parameters)`29: Requires `parameters.port` and `parameters.action` where action is `allow` or `deny`.30 31- `restart_service(resource_id)`32: Use only after causal evidence confirms target.33 34- `submit_solution`35: Useful for explicit closure checks; unresolved hard submissions are penalized and continue.36 37## Task Playbooks38 39### Easy40 41Objective:42- Restore web access by allowing port `80` on `sg-web`.43 44Strong path:451. `list_resources`462. `update_security_group("sg-web", {"port": 80, "action": "allow"})`47 48Typical outcome:49- 2 steps, score around `0.78`50 51### Medium52 53Objective:54- Restore API to DB connectivity by allowing port `5432` on `sg-db`.55 56Strong path:571. `list_resources`582. `view_logs("i-api")`593. `query_metadata({"ip_address": "10.0.4.5"})`604. `update_security_group("sg-db", {"port": 5432, "action": "allow"})`61 62Guardrail:63- Applying SG changes before logs + metadata lookup is penalized and does not resolve the incident.64 65### Hard66 67Objective:68- Recover checkout flow by identifying the failing upstream and restarting `i-web2` safely.69 70Strong path:711. `list_resources`722. `view_logs("lb-main")`733. `query_metadata({"ip_address": "10.0.8.22"})`744. `describe_resource("i-web2")` or `view_logs("i-web2")`755. `restart_service("i-web2")`76 77Guardrails:78- Restarting `i-web1` is penalized.79- Restarting `i-web2` without investigation is penalized.80- If unresolved after step 8, `lb-external` also fails (cascading failure).81 82## Reading Metadata83 84Watch these response fields each step:85 86- `system_health_status`: `CRITICAL` / `DEGRADED` / `HEALTHY`87- `done`: episode ended or still running88- `reward`: immediate signal after action cost and shaping89- `metadata.resolved`: authoritative success flag90- `metadata.termination_reason`: why episode ended91- `metadata.reward_breakdown`: transparent reward events for grader/debug inspection92 93## Submission Contract Reminder94 95`inference.py` must:96 97- use OpenAI client98- read `API_BASE_URL`, `MODEL_NAME`, `HF_TOKEN`99- emit strict stdout markers: `[START]`, `[STEP]`, `[END]`100