openenv.yaml

spec_version: 1
name: netweaver_sre
type: space
runtime: fastapi
app: server.app:app
port: 7860

tasks:

  # ── EASY ──────────────────────────────────────────────────────────────────

  - id: netweaver_sre_t01
    difficulty: easy
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t01
    description: >
      Node Offline Triage. Read hardware_logs to find the offline node.
      Remediate with DRAIN_TRAFFIC on the identified node.
      Scored on: diagnosis (40%) + resolution (40%) + best practice (20%).

  - id: netweaver_sre_t02
    difficulty: easy
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t02
    description: >
      Stale DNS Cache. Read hardware_logs to identify the node.
      Issue CLEAR_DNS_CACHE on the affected node.
      Scored on: diagnosis (40%) + resolution (40%) + best practice (20%).

  - id: netweaver_sre_t03
    difficulty: easy
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t03
    description: >
      OOM Crash. Read hardware_logs for the crashed service name.
      Issue RESTART_SERVICE on that service.
      Scored on: diagnosis (40%) + resolution (40%) + best practice (20%).

  - id: netweaver_sre_t04
    difficulty: easy
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t04
    description: >
      TLS Certificate Expiry. Read hardware_logs for the affected node.
      Issue RENEW_CERTIFICATE on that node.
      Scored on: diagnosis (40%) + resolution (40%) + best practice (20%).

  - id: netweaver_sre_t05
    difficulty: easy
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t05
    description: >
      Disk Space Full. Read hardware_logs for the full node.
      Issue CLEAR_TEMP_FILES on that node.
      Scored on: diagnosis (40%) + resolution (40%) + best practice (20%).

  - id: netweaver_sre_t06
    difficulty: easy
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t06
    description: >
      Unhealthy Pod in CrashLoopBackOff. Read hardware_logs for the pod name.
      Issue RESTART_POD on that pod.
      Scored on: diagnosis (40%) + resolution (40%) + best practice (20%).

  - id: netweaver_sre_t07
    difficulty: easy
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t07
    description: >
      Zombie Process filling PID table. Read hardware_logs for the node.
      Issue KILL_ZOMBIE_PROCESS on that node.
      Scored on: diagnosis (40%) + resolution (40%) + best practice (20%).

  # ── MEDIUM ────────────────────────────────────────────────────────────────

  - id: netweaver_sre_t08
    difficulty: medium
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t08
    description: >
      PFC Buffer Congestion. Read queue_depths for the switch near 99.9.
      Issue TUNE_PFC_THRESHOLD on that switch with a numeric value (1000-9000).
      Grader checks correct target AND value in valid range.
      Efficiency penalty applied for steps beyond ideal (3).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t09
    difficulty: medium
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t09
    description: >
      Power Throttling. Read hardware_logs for node and recommended wattage.
      Issue ADJUST_POWER_CAP on that node with the watt value.
      Efficiency penalty applied for steps beyond ideal (3).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t10
    difficulty: medium
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t10
    description: >
      BGP Route Flapping. Read hardware_logs for router and AS number.
      Issue MITIGATE_ROUTE_FLAP on that router with the AS number as value.
      Efficiency penalty applied for steps beyond ideal (3).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t11
    difficulty: medium
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t11
    description: >
      Jumbo Frame Packet Drop (MTU mismatch). Read queue_depths/logs for switch.
      Issue INCREASE_MTU on that switch with value exactly 9000.
      Efficiency penalty applied for steps beyond ideal (3).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t12
    difficulty: medium
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t12
    description: >
      DDoS API flood. Read hardware_logs for the gateway name.
      Issue SET_RATE_LIMIT on the gateway with a numeric req/s value.
      Efficiency penalty applied for steps beyond ideal (3).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t13
    difficulty: medium
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t13
    description: >
      DB Connection Pool Exhaustion. Read hardware_logs for DB target.
      Issue SCALE_CONN_POOL on that DB with the new pool size as value.
      Efficiency penalty applied for steps beyond ideal (3).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t14
    difficulty: medium
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t14
    description: >
      CPU Context Switch Storm. Read hardware_logs for node and thread count.
      Issue PIN_CPU_THREADS on that node with thread count as value.
      Efficiency penalty applied for steps beyond ideal (3).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  # ── HARD ──────────────────────────────────────────────────────────────────

  - id: netweaver_sre_t15
    difficulty: hard
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t15
    description: >
      Silent NaN Contagion. Read gradient_variances for ranks with value -1.0.
      Multi-step: (1) RUN_MINI_ITERATION on affected cluster to isolate,
      then (2) DRAIN_TRAFFIC on same cluster to remove from pool.
      Both commands required for full resolution score.
      Efficiency penalty applied for steps beyond ideal (4).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t16
    difficulty: hard
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t16
    description: >
      Broadcast Storm. Read queue_depths for the switch at 99.9.
      Issue ISOLATE_BROADCAST_STORM on that exact switch.
      Efficiency penalty applied for steps beyond ideal (3).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t17
    difficulty: hard
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t17
    description: >
      GPU Memory Leak. Read gpu_memory_usage for the cluster index near 1.0.
      Issue RESTART_GPU_DAEMON on cluster_<that index>.
      Efficiency penalty applied for steps beyond ideal (3).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t18
    difficulty: hard
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t18
    description: >
      Full Cluster Deadlock. All arrays frozen at 0.0 — distributed deadlock.
      Issue ISSUE_GLOBAL_ROLLBACK with the exact cluster target from logs.
      Efficiency penalty applied for steps beyond ideal (3).
      Scored on: diagnosis (40%) + resolution (40%) + best practice (20%).

  - id: netweaver_sre_t19
    difficulty: hard
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t19
    description: >
      Network Partition. Read queue_depths for the split (0.01 vs 99.9).
      Issue REBOOT_LEAF_SWITCHES on the pod with the maxed entry.
      Efficiency penalty applied for steps beyond ideal (4).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t20
    difficulty: hard
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t20
    description: >
      Corrupt DB Block. Read logs for the randomized DB cluster target.
      Issue PURGE_CORRUPT_BLOCK on that exact DB target.
      Efficiency penalty applied for steps beyond ideal (4).
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t21
    difficulty: hard
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t21
    description: >
      Cascading Failure Chain. A power fault triggers a CPU context switch storm
      which overflows the connection pool. Requires 3 sequential remediations:
      (1) ADJUST_POWER_CAP on the throttled node,
      (2) PIN_CPU_THREADS on the same node,
      (3) SCALE_CONN_POOL on the database.
      Order matters. Partial credit for each completed step.
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).

  - id: netweaver_sre_t22
    difficulty: hard
    grader:
      type: deterministic
      endpoint: /grader
      task_level: t22
    description: >
      Silent Gradient Poisoning with Network Amplification.
      NaN is corrupting gradient sync AND a broadcast storm is amplifying the corrupted
      gradients across the fabric. Requires:
      (1) RUN_MINI_ITERATION to isolate NaN source cluster,
      (2) ISOLATE_BROADCAST_STORM on the storm switch,
      (3) DRAIN_TRAFFIC on the poisoned cluster.
      All 3 commands required in this order for full resolution.
      Scored on: diagnosis (40%) + resolution+efficiency (40%) + best practice (20%).