Fix cloud multi-nodes

* Copy ssh key to allow connections from master to workers
* Use local ip for manager's ip such that workers can find it and connect to it
* Fix incompatibility between pandas and numpy 2.0.0

.github/workflows/cloud-ci.yml

@@ -14,11 +14,12 @@ jobs:
   cloud-tests:
     strategy:
       fail-fast: true
+      max-parallel: 1
       matrix:
+        system: [1x_gpu, 2x_gpu]
         include:
           - arch: cuda
             exclude: "no-cuda"
-            run_on: azure__a100
           # - arch: rocm
           #   exclude : "no-rocm"
 
@@ -27,7 +28,7 @@ jobs:
 
     # Cancel previous jobs if a new version was pushed
     concurrency:
-      group: "${{ github.ref }}-${{ matrix.arch }}-${{ matrix.run_on }}"
+      group: "${{ github.ref }}-${{ matrix.arch }}-${{ matrix.system }}"
       cancel-in-progress: true
 
     defaults:
@@ -36,7 +37,7 @@ jobs:
 
     env:
       MILABENCH_CONFIG: "config/standard.yaml"
-      MILABENCH_SYSTEM: "config/cloud-system.yaml"
+      MILABENCH_SYSTEM: "config/cloud-multinodes-system.yaml"
       MILABENCH_BASE: "output"
       MILABENCH_ARGS: ""
       MILABENCH_DASH: "no"
@@ -90,23 +91,51 @@ jobs:
 
       - name: setup cloud
         run: |
+          case "${{ matrix.system }}" in
+            "1x_gpu")
+              export MILABENCH_SYSTEM="config/cloud-system.yaml"
+              export RUN_ON="azure__a100"
+              export SELECT=
+              export EXCLUDES=--exclude dinov2-giant-gpus,lightning-gpus,resnet152-ddp-gpus,llm-full-mp-gpus,llm-full-mp-nodes,llm-lora-ddp-gpus,llm-lora-ddp-nodes,llm-lora-mp-gpus,llm-lora-single
+              ;;
+            "2x_gpu")
+              export MILABENCH_SYSTEM="config/cloud-system.yaml"
+              export RUN_ON="azure__a100_x2"
+              export SELECT=dinov2-giant-gpus,lightning-gpus,resnet152-ddp-gpus
+              export EXCLUDES=
+              ;;
+            "2x_node")
+              export MILABENCH_SYSTEM="config/cloud-multinodes-system.yaml"
+              export RUN_ON="azure__a100"
+              export SELECT=
+              export EXCLUDES=
+              ;;
+            *)
+              exit 1
+              ;;
+          esac
+
           poetry run milabench cloud \
             --setup \
-            --run-on ${{ matrix.run_on }} \
-            --system "$MILABENCH_SYSTEM" >$MILABENCH_SYSTEM.${{ matrix.run_on }}
-          echo "MILABENCH_SYSTEM=$MILABENCH_SYSTEM.${{ matrix.run_on }}" >>$GITHUB_ENV
+            --run-on $RUN_ON \
+            --system "$MILABENCH_SYSTEM" >$MILABENCH_SYSTEM.$RUN_ON
+
+          echo "RUN_ON=$RUN_ON" >>$GITHUB_ENV
+          echo "SELECT=$SELECT" >>$GITHUB_ENV
+          echo "EXCLUDES=$EXCLUDES" >>$GITHUB_ENV
+          echo "MILABENCH_SYSTEM=$MILABENCH_SYSTEM.$RUN_ON" >>$GITHUB_ENV
 
       - name: install benchmarks
         run: |
-          poetry run milabench install --variant ${{ matrix.arch }}
+          poetry run milabench install --variant ${{ matrix.arch }} $SELECT $EXCLUDES
 
       - name: prepare benchmarks
         run: |
-          poetry run milabench prepare
+          poetry run milabench prepare $SELECT $EXCLUDES
 
       - name: run benchmarks
         run: |
-          poetry run milabench run
+          poetry run milabench run $SELECT $EXCLUDES
 
       - name: Summary
         run: |
@@ -118,6 +147,11 @@ jobs:
         env:
           GITHUB_TOKEN: ${{ github.token }}
 
+      - name: DEBUG state file
+        if: always()
+        run: |
+          cat /tmp/milabench/covalent_venv/lib/python*/site-packages/covalent_azure_plugin/infra/*.tfstate
+
       - name: teardown cloud
         if: always()
         run: |
@@ -127,10 +161,10 @@ jobs:
           fi
           poetry run milabench cloud \
             --teardown \
-            --run-on ${{ matrix.run_on }} \
+            --run-on $RUN_ON \
             --all
 
-      - name: debug logs
+      - name: DEBUG logs
         if: always()
         run: |
           cat ~/.cache/covalent/covalent_ui.log

config/cloud-multinodes-system.yaml

@@ -5,6 +5,7 @@ system:
     - name: manager
       # Use 1.1.1.1 as an ip placeholder
       ip: 1.1.1.1
+      port: 5000
       # Use this node as the master node or not
       main: true
       # User to use in remote milabench operations
@@ -21,11 +22,14 @@ system:
       username: ubuntu
       size: Standard_NC24ads_A100_v4
       location: eastus2
+      disk_size: 512
     azure__a100_x2:
       username: ubuntu
       size: Standard_NC48ads_A100_v4
       location: eastus2
+      disk_size: 512
     azure__a10_x2:
       username: ubuntu
       size: Standard_NV72ads_A10_v5
       location: eastus2
+      disk_size: 512

config/cloud-system.yaml

@@ -16,11 +16,19 @@ system:
       username: ubuntu
       size: Standard_NC24ads_A100_v4
       location: eastus2
+      disk_size: 512
     azure__a100_x2:
       username: ubuntu
       size: Standard_NC48ads_A100_v4
       location: eastus2
+      disk_size: 512
+    azure__a10:
+      username: ubuntu
+      size: Standard_NV36ads_A10_v5
+      location: eastus2
+      disk_size: 512
     azure__a10_x2:
       username: ubuntu
       size: Standard_NV72ads_A10_v5
       location: eastus2
+      disk_size: 512

config/examples/test.yaml

@@ -7,18 +7,18 @@ _defaults:
 
 test:
   inherits: _defaults
-  group: test_remote
-  install_group: test_remote
-  definition: ../../benchmarks/_template
+  group: simple
+  install_group: test
+  definition: ../../benchmarks/_templates/simple
   plan:
     method: njobs
     n: 1
 
 testing:
   inherits: _defaults
-  definition: ../../benchmarks/_template
-  group: test_remote_2
-  install_group: test_remote_2
+  definition: ../../benchmarks/_templates/stdout
+  group: stdout
+  install_group: test
   plan:
     method: njobs
     n: 1

milabench/commands/__init__.py

@@ -939,6 +939,13 @@ def _get_main_and_workers(self):
     def _argv(self, **_) -> List:
         manager, nodes = self._get_main_and_workers()
 
+        # Find local ip such that workers can connect to the port
+        for manager_ip in manager["ipaddrlist"]:
+            if ":" in manager_ip or manager_ip == "127.0.0.1":
+                continue
+            if all(str.isnumeric(n) for n in manager_ip.split(".")):
+                break
+
         num_machines = max(1, len(nodes) + 1)
 
         # Cant do that maybe this run is constrained
@@ -976,9 +983,9 @@ def _argv(self, **_) -> List:
             f"--machine_rank={self.rank}",
             f"--num_machines={num_machines}",
             *deepspeed_argv,
-            f"--gradient_accumulation_steps={self.pack.config.get('gradient_accumulation_steps', 1)}",
-            f"--num_cpu_threads_per_process={cpu_per_process}",
-            f"--main_process_ip={manager['ip']}",
+            f"--gradient_accumulation_steps={self.pack.config['gradient_accumulation_steps']}",
+            f"--num_cpu_threads_per_process={self.pack.config['argv']['--cpus_per_gpu']}",
+            f"--main_process_ip={manager_ip}",
             f"--main_process_port={manager['port']}",
             f"--num_processes={nproc}",
             *self.accelerate_argv,

milabench/config.py

@@ -150,6 +150,7 @@ def build_config(*config_files):
     for layer in _config_layers(config_files):
         all_configs = merge(all_configs, layer)
 
+    all_configs.setdefault("*", {})
     all_configs["*"]["hash"] = compute_config_hash(all_configs)
 
     all_configs = build_matrix_bench(all_configs)

milabench/remote.py

@@ -2,16 +2,11 @@
 import os
 import sys
 
-import yaml
-
-from milabench.fs import XPath
-
 from . import ROOT_FOLDER
 from .commands import (
     CmdCommand,
     Command,
     ListCommand,
-    SCPCommand,
     SequenceCommand,
     SSHCommand,
     VoidCommand,

milabench/scripts/covalent/__main__.py

@@ -89,33 +89,80 @@ def _popen(cmd, *args, _env=None, **kwargs):
     return_code = 0
     try:
         if args.setup:
-            dispatch_id = ct.dispatch(
-                ct.lattice(executor.get_connection_attributes), disable_run=False
-            )()
-
-            result = ct.get_result(dispatch_id=dispatch_id, wait=True).result
+            result = ct.dispatch_sync(
+                ct.lattice(executor.get_connection_attributes)
+            )().result
 
             assert result and result[0]
 
             all_connection_attributes, _ = result
+            master_host:str = None
             for hostname, connection_attributes in all_connection_attributes.items():
                 print(f"hostname::>{hostname}")
                 for attribute,value in connection_attributes.items():
                     if attribute == "hostname":
                         continue
                     print(f"{attribute}::>{value}")
 
-        if argv:
-            dispatch_id = ct.dispatch(
-                ct.lattice(
-                    lambda:ct.electron(_popen, executor=executor)(argv)
-                ),
-                disable_run=False
-            )()
+                master_host = master_host or hostname
+
+            if len(all_connection_attributes) > 1:
+                # Add master node to known host to avoid unknown host error
+                # The authenticity of host '[hostname] ([IP address])' can't be established.
+                new_host = subprocess.run(
+                    ["ssh-keyscan", master_host],
+                    stdout=subprocess.PIPE,
+                    check=True
+                ).stdout.decode("utf8")
+                known_hosts = pathlib.Path("~/.ssh/known_hosts").expanduser()
+                with known_hosts.open("at") as _f:
+                    _f.write(new_host)
+
+                # Add ssh file to master node to allow connections to worker
+                # nodes
+                ssh_key_file = all_connection_attributes[master_host]["ssh_key_file"]
+                fn = pathlib.Path(ssh_key_file)
+                result = ct.dispatch_sync(
+                    ct.lattice(executor.cp_to_remote)
+                )(f".ssh/{fn.name.split('.')[0]}", str(fn))
+
+                assert result.status == ct.status.COMPLETED
 
-            result = ct.get_result(dispatch_id=dispatch_id, wait=True).result
+        if argv:
+            result = ct.dispatch_sync(
+                ct.lattice(executor.list_running_instances)
+            )().result
+
+            assert result
+
+            dispatch_ids = set()
+            for connection_attributes in result.get(
+                (executor.state_prefix, executor.state_id),
+                {"env": None}
+            ).values():
+                kwargs = {
+                    **_get_executor_kwargs(args),
+                    **connection_attributes
+                }
+                del kwargs["env"]
+
+                _executor:ct.executor.BaseExecutor = executor_cls(**kwargs)
+
+                dispatch_ids.add(
+                    ct.dispatch(
+                        ct.lattice(
+                            lambda:ct.electron(_popen, executor=_executor)(argv)
+                        ),
+                        disable_run=False
+                    )()
+                )
+
+            for dispatch_id in dispatch_ids:
+                result = ct.get_result(dispatch_id=dispatch_id, wait=True).result
+
+                _return_code, _, _ = result if result is not None else (1, "", "")
+                return_code = return_code or _return_code
 
-            return_code, _, _ = result if result is not None else (1, "", "")
     finally:
         if args.teardown:
             result = executor.stop_cloud_instance().result

milabench/system.py

@@ -258,6 +258,10 @@ def _resolve_ip(ip):
     if not offline:
         # Resolve the IP
         try:
+            # Workaround error with `gethostbyaddr` on azure DNS (like
+            # `inmako.eastus2.cloudapp.azure.com`). A proper fix might be a
+            # correct network config in terraform.
+            # socket.herror: [Errno 1] Unknown host
             hostname, aliaslist, ipaddrlist = socket.gethostbyname_ex(ip)
             lazy_raise = None
 

pyproject.toml

@@ -26,7 +26,9 @@ blessed = "^1.19.1"
 pathspec = "^0.9.0"
 cp-template = "^0.3.0"
 pandas = ">=1.4.2"
-numpy = ">=1.23.0,<2.0.0"
+# Work around for compatibility issue between numpy 2.0.0 and pandas
+# https://github.com/numpy/numpy/issues/26710
+numpy = "^1.23.0"
 pynvml = "^11.4.1"
 tqdm = "^4.64.1"
 pip-tools = "^7.4.1"