deploy/helm/health_checks/nccl_healthcheck/a3ultra.yaml

# Copyright 2024 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

job:
  base_name: "chs-hc"
  # guid: "xckd"  # Can specify a GUID if desired. Otherwise, a random GUID will be generated.
  # check_time: "1590303600"  # Will automatically be set if not given
health_check:
  name: "nccl"
  image:
    repo: "us-docker.pkg.dev/gce-ai-infra/health-check/nccl-healthcheck"
    tag: "a3-ultragpu-8g_4.4.0"
    pull_policy: "Always"
  env:
    # A3+ instance type is a3-megagpu-8g
    INSTANCE_TYPE: "a3-ultragpu-8g"
    ENABLE_TWO_PASS_STRATEGY: "true"
    HEALTH_VALIDITY_HOURS: "24"
    DRY_RUN: "true"
    START_MESSAGE_SIZE: "2G"
    END_MESSAGE_SIZE: "8G"
    TEST_ITERATIONS: "3"
    NHOSTS: "2"
    nr: "8"
    # Specific to A3+
    LD_LIBRARY_PATH: /usr/local/gib/lib64:/usr/local/nvidia/lib64
    BANDWIDTH_THRESHOLD: "250"
    USE_TCPX: "false"
    USE_FASTRAK: "false"
    # Note A3+ (a3-megagpu-8g) has no UNIX_CLIENT_PREFIX
    NCCL_LIB_DIR: /usr/local/nvidia/lib64
    NCCL_FASTRAK_USE_SNAP: "1"
    NCCL_FASTRAK_ENABLE_CONTROL_CHANNEL: "0"
    NCCL_FASTRAK_NUM_FLOWS: "2"
    NCCL_DEBUG: "INFO"
    NCCL_DEBUG_SUBSYS: "INIT,NET"
  volumeMounts:
  - name: usr-local-gib
    mountPath: /usr/local/gib
  - name: library-dir-host
    mountPath: /usr/local/nvidia
  - name: shared-memory
    mountPath: /dev/shm
  - name: varlog
    mountPath: /var/log
# Configuration specific to a3ultra
volumes:
- name: library-dir-host
  hostPath:
    path: /home/kubernetes/bin/nvidia
- name: shared-memory
  emptyDir:
    medium: "Memory"
    sizeLimit: 250Gi
- name: usr-local-gib
  hostPath:
    path: /home/kubernetes/bin/gib
- name: run-sshd
  emptyDir: {}
- name: varlog
  emptyDir: {}
- name: fluentbit-key
  secret:
    secretName: fluentbit-key
    optional: true
initContainers:
  nccl_plugin_installer:
    name: "nccl-plugin-installer"
    image: "us-docker.pkg.dev/gce-ai-infra/gpudirect-gib/nccl-plugin-gib:v1.0.2"
    imagePullPolicy: "Always"
    volumeMounts:
      name: "usr-local-gib"
      mountPath: "/usr/local/gib"