Skip to content

GitLab

Unverified Commit a5049f0f authored by Rohan Varma's avatar Rohan Varma Committed by GitHub
Browse files

fix: inject TRTLLM_USE_UCX_KVCACHE env var in mpirun (#4609)

parent fcb91e4b
Show whitespace changes
Inline Side-by-side
Showing with 11 additions and 11 deletions
deploy/cloud/operator/internal/dynamo/backend_trtllm.go
View file @ a5049f0f
...@@ -224,7 +224,7 @@ func getCommonTRTLLMEnvVars() map[string]bool { ...@@ -224,7 +224,7 @@ func getCommonTRTLLMEnvVars() map[string]bool {
"CUDA_VISIBLE_DEVICES": true, "MODEL_PATH": true, "HF_TOKEN": true, "HUGGING_FACE_HUB_TOKEN": true, "HF_ENDPOINT": true, "CUDA_VISIBLE_DEVICES": true, "MODEL_PATH": true, "HF_TOKEN": true, "HUGGING_FACE_HUB_TOKEN": true, "HF_ENDPOINT": true,
"TOKENIZERS_PARALLELISM": true, "NCCL_DEBUG": true, "NCCL_IB_DISABLE": true, "NCCL_P2P_DISABLE": true, "TOKENIZERS_PARALLELISM": true, "NCCL_DEBUG": true, "NCCL_IB_DISABLE": true, "NCCL_P2P_DISABLE": true,
"TENSORRT_LLM_CACHE_DIR": true, "HF_HOME": true, "TRANSFORMERS_CACHE": true, "HF_DATASETS_CACHE": true, "TENSORRT_LLM_CACHE_DIR": true, "HF_HOME": true, "TRANSFORMERS_CACHE": true, "HF_DATASETS_CACHE": true,
"PATH": true, "LD_LIBRARY_PATH": true, "PYTHONPATH": true, "HOME": true, "USER": true, "PATH": true, "LD_LIBRARY_PATH": true, "PYTHONPATH": true, "HOME": true, "USER": true, "TRTLLM_USE_UCX_KVCACHE": true,
} }
} }
......
deploy/cloud/operator/internal/dynamo/backend_trtllm_test.go
View file @ a5049f0f
...@@ -61,7 +61,7 @@ func TestTRTLLMBackend_UpdateContainer(t *testing.T) { ...@@ -61,7 +61,7 @@ func TestTRTLLMBackend_UpdateContainer(t *testing.T) {
{Name: mpiRunSecretName, MountPath: "/ssh-pk", ReadOnly: true}, {Name: mpiRunSecretName, MountPath: "/ssh-pk", ReadOnly: true},
}, },
expectedCommand: []string{"/bin/sh", "-c"}, expectedCommand: []string{"/bin/sh", "-c"},
expectedArgs: []string{"mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 6 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-wkr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-wkr-1.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x OMPI_MCA_orte_keep_fqdn_hostnames -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'trtllm-llmapi-launch python3 --model test'"}, expectedArgs: []string{"mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 6 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-wkr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-wkr-1.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x OMPI_MCA_orte_keep_fqdn_hostnames -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x TRTLLM_USE_UCX_KVCACHE -x USER bash -c 'trtllm-llmapi-launch python3 --model test'"},
expectedEnv: []corev1.EnvVar{ expectedEnv: []corev1.EnvVar{
{Name: "OMPI_MCA_orte_keep_fqdn_hostnames", Value: "1"}, {Name: "OMPI_MCA_orte_keep_fqdn_hostnames", Value: "1"},
}, },
...@@ -115,7 +115,7 @@ func TestTRTLLMBackend_UpdateContainer(t *testing.T) { ...@@ -115,7 +115,7 @@ func TestTRTLLMBackend_UpdateContainer(t *testing.T) {
{Name: mpiRunSecretName, MountPath: "/ssh-pk", ReadOnly: true}, {Name: mpiRunSecretName, MountPath: "/ssh-pk", ReadOnly: true},
}, },
expectedCommand: []string{"/bin/sh", "-c"}, expectedCommand: []string{"/bin/sh", "-c"},
expectedArgs: []string{"mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && TIMEOUT=300; START_TIME=$(date +%s); for worker in $(echo \"$LWS_LEADER_ADDRESS,$(echo \"$LWS_LEADER_ADDRESS\" | sed 's/\\./-1\\./')\" | tr ',' ' '); do echo \"Waiting for DNS: $worker\"; until getent hosts $worker >/dev/null 2>&1; do CURRENT_TIME=$(date +%s); if [ $((CURRENT_TIME - START_TIME)) -gt $TIMEOUT ]; then echo \"ERROR: Timeout waiting for DNS: $worker\"; exit 1; fi; echo \"DNS not ready for $worker, retrying...\"; sleep 2; done; echo \"✓ DNS resolved: $worker\"; done; echo \"All workers DNS ready\" && mpirun --allow-run-as-root --oversubscribe -n 2 -H $LWS_LEADER_ADDRESS,$(echo \"$LWS_LEADER_ADDRESS\" | sed 's/\\./-1\\./') --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x OMPI_MCA_orte_keep_fqdn_hostnames -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'trtllm-llmapi-launch python3 --model test'"}, expectedArgs: []string{"mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && TIMEOUT=300; START_TIME=$(date +%s); for worker in $(echo \"$LWS_LEADER_ADDRESS,$(echo \"$LWS_LEADER_ADDRESS\" | sed 's/\\./-1\\./')\" | tr ',' ' '); do echo \"Waiting for DNS: $worker\"; until getent hosts $worker >/dev/null 2>&1; do CURRENT_TIME=$(date +%s); if [ $((CURRENT_TIME - START_TIME)) -gt $TIMEOUT ]; then echo \"ERROR: Timeout waiting for DNS: $worker\"; exit 1; fi; echo \"DNS not ready for $worker, retrying...\"; sleep 2; done; echo \"✓ DNS resolved: $worker\"; done; echo \"All workers DNS ready\" && mpirun --allow-run-as-root --oversubscribe -n 2 -H $LWS_LEADER_ADDRESS,$(echo \"$LWS_LEADER_ADDRESS\" | sed 's/\\./-1\\./') --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x OMPI_MCA_orte_keep_fqdn_hostnames -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x TRTLLM_USE_UCX_KVCACHE -x USER bash -c 'trtllm-llmapi-launch python3 --model test'"},
expectedEnv: []corev1.EnvVar{ expectedEnv: []corev1.EnvVar{
{Name: "OMPI_MCA_orte_keep_fqdn_hostnames", Value: "1"}, {Name: "OMPI_MCA_orte_keep_fqdn_hostnames", Value: "1"},
}, },
...@@ -564,7 +564,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -564,7 +564,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
}, },
initialArgs: []string{"python3", "--model", "test"}, initialArgs: []string{"python3", "--model", "test"},
initialCommand: []string{}, initialCommand: []string{},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 6 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-wkr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-wkr-1.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'trtllm-llmapi-launch python3 --model test'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 6 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-wkr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-service-wkr-1.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x TRTLLM_USE_UCX_KVCACHE -x USER bash -c 'trtllm-llmapi-launch python3 --model test'",
}, },
{ {
name: "Leader with command and no GPU resources", name: "Leader with command and no GPU resources",
...@@ -574,7 +574,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -574,7 +574,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
component: &v1alpha1.DynamoComponentDeploymentSharedSpec{}, component: &v1alpha1.DynamoComponentDeploymentSharedSpec{},
initialArgs: []string{}, initialArgs: []string{},
initialCommand: []string{"python", "-m", "worker"}, initialCommand: []string{"python", "-m", "worker"},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && TIMEOUT=300; START_TIME=$(date +%s); for worker in $(echo \"$LWS_LEADER_ADDRESS,$(echo \"$LWS_LEADER_ADDRESS\" | sed 's/\\./-1\\./')\" | tr ',' ' '); do echo \"Waiting for DNS: $worker\"; until getent hosts $worker >/dev/null 2>&1; do CURRENT_TIME=$(date +%s); if [ $((CURRENT_TIME - START_TIME)) -gt $TIMEOUT ]; then echo \"ERROR: Timeout waiting for DNS: $worker\"; exit 1; fi; echo \"DNS not ready for $worker, retrying...\"; sleep 2; done; echo \"✓ DNS resolved: $worker\"; done; echo \"All workers DNS ready\" && mpirun --allow-run-as-root --oversubscribe -n 0 -H $LWS_LEADER_ADDRESS,$(echo \"$LWS_LEADER_ADDRESS\" | sed 's/\\./-1\\./') --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'trtllm-llmapi-launch python -m worker'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && TIMEOUT=300; START_TIME=$(date +%s); for worker in $(echo \"$LWS_LEADER_ADDRESS,$(echo \"$LWS_LEADER_ADDRESS\" | sed 's/\\./-1\\./')\" | tr ',' ' '); do echo \"Waiting for DNS: $worker\"; until getent hosts $worker >/dev/null 2>&1; do CURRENT_TIME=$(date +%s); if [ $((CURRENT_TIME - START_TIME)) -gt $TIMEOUT ]; then echo \"ERROR: Timeout waiting for DNS: $worker\"; exit 1; fi; echo \"DNS not ready for $worker, retrying...\"; sleep 2; done; echo \"✓ DNS resolved: $worker\"; done; echo \"All workers DNS ready\" && mpirun --allow-run-as-root --oversubscribe -n 0 -H $LWS_LEADER_ADDRESS,$(echo \"$LWS_LEADER_ADDRESS\" | sed 's/\\./-1\\./') --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x TRTLLM_USE_UCX_KVCACHE -x USER bash -c 'trtllm-llmapi-launch python -m worker'",
}, },
{ {
name: "Leader with both command and args (shell command - args take precedence)", name: "Leader with both command and args (shell command - args take precedence)",
...@@ -590,7 +590,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -590,7 +590,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
}, },
initialArgs: []string{"launch", "--config", "test.yaml"}, initialArgs: []string{"launch", "--config", "test.yaml"},
initialCommand: []string{"sh", "-c"}, initialCommand: []string{"sh", "-c"},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'trtllm-llmapi-launch launch --config test.yaml'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x TRTLLM_USE_UCX_KVCACHE -x USER bash -c 'trtllm-llmapi-launch launch --config test.yaml'",
}, },
{ {
name: "Leader with python command and args (combined)", name: "Leader with python command and args (combined)",
...@@ -606,7 +606,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -606,7 +606,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
}, },
initialArgs: []string{"-m", "dynamo.trtllm", "--model-path", "test"}, initialArgs: []string{"-m", "dynamo.trtllm", "--model-path", "test"},
initialCommand: []string{"python3"}, initialCommand: []string{"python3"},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'trtllm-llmapi-launch python3 -m dynamo.trtllm --model-path test'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x TRTLLM_USE_UCX_KVCACHE -x USER bash -c 'trtllm-llmapi-launch python3 -m dynamo.trtllm --model-path test'",
}, },
{ {
name: "Leader with python module command and separate args", name: "Leader with python module command and separate args",
...@@ -622,7 +622,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -622,7 +622,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
}, },
initialArgs: []string{"--model-path", "Qwen/Qwen3-0.6B", "--served-model-name", "Qwen/Qwen3-0.6B", "--disaggregation-mode", "prefill"}, initialArgs: []string{"--model-path", "Qwen/Qwen3-0.6B", "--served-model-name", "Qwen/Qwen3-0.6B", "--disaggregation-mode", "prefill"},
initialCommand: []string{"python3", "-m", "dynamo.trtllm"}, initialCommand: []string{"python3", "-m", "dynamo.trtllm"},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'trtllm-llmapi-launch python3 -m dynamo.trtllm --model-path Qwen/Qwen3-0.6B --served-model-name Qwen/Qwen3-0.6B --disaggregation-mode prefill'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x TRTLLM_USE_UCX_KVCACHE -x USER bash -c 'trtllm-llmapi-launch python3 -m dynamo.trtllm --model-path Qwen/Qwen3-0.6B --served-model-name Qwen/Qwen3-0.6B --disaggregation-mode prefill'",
}, },
{ {
name: "Leader with absolute path python command", name: "Leader with absolute path python command",
...@@ -638,7 +638,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -638,7 +638,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
}, },
initialArgs: []string{"-m", "dynamo.trtllm", "--model-path", "test"}, initialArgs: []string{"-m", "dynamo.trtllm", "--model-path", "test"},
initialCommand: []string{"/usr/bin/python3.8"}, initialCommand: []string{"/usr/bin/python3.8"},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'trtllm-llmapi-launch /usr/bin/python3.8 -m dynamo.trtllm --model-path test'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x TRTLLM_USE_UCX_KVCACHE -x USER bash -c 'trtllm-llmapi-launch /usr/bin/python3.8 -m dynamo.trtllm --model-path test'",
}, },
{ {
name: "Leader with all environment variables forwarded", name: "Leader with all environment variables forwarded",
...@@ -654,7 +654,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -654,7 +654,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
}, },
initialArgs: []string{"serve", "--model", "test"}, initialArgs: []string{"serve", "--model", "test"},
initialCommand: []string{}, initialCommand: []string{},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'trtllm-llmapi-launch serve --model test'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x TRTLLM_USE_UCX_KVCACHE -x USER bash -c 'trtllm-llmapi-launch serve --model test'",
}, },
{ {
name: "Leader with overlapping environment variables (deduplication test)", name: "Leader with overlapping environment variables (deduplication test)",
...@@ -670,7 +670,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) { ...@@ -670,7 +670,7 @@ func TestTRTLLMBackend_setupLeaderContainer(t *testing.T) {
}, },
initialArgs: []string{"serve", "--model", "test"}, initialArgs: []string{"serve", "--model", "test"},
initialCommand: []string{}, initialCommand: []string{},
expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x CUSTOM_VAR -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x USER bash -c 'trtllm-llmapi-launch serve --model test'", expected: "mkdir -p ~/.ssh && ls -la /ssh-pk/ && cp /ssh-pk/private.key ~/.ssh/id_rsa && cp /ssh-pk/private.key.pub ~/.ssh/id_rsa.pub && cp /ssh-pk/private.key.pub ~/.ssh/authorized_keys && chmod 600 ~/.ssh/id_rsa ~/.ssh/authorized_keys && chmod 644 ~/.ssh/id_rsa.pub ~/.ssh/authorized_keys && printf 'Host *\\nIdentityFile ~/.ssh/id_rsa\\nStrictHostKeyChecking no\\nPort 2222\\n' > ~/.ssh/config && mpirun --allow-run-as-root --oversubscribe -n 2 -H $(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-ldr-0.$(GROVE_HEADLESS_SERVICE),$(GROVE_PCSG_NAME)-$(GROVE_PCSG_INDEX)-test-wkr-0.$(GROVE_HEADLESS_SERVICE) --mca pml ob1 --mca plm_rsh_args \"-p 2222 -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" -x CUDA_VISIBLE_DEVICES -x CUSTOM_VAR -x HF_DATASETS_CACHE -x HF_ENDPOINT -x HF_HOME -x HF_TOKEN -x HOME -x HUGGING_FACE_HUB_TOKEN -x LD_LIBRARY_PATH -x MODEL_PATH -x NCCL_DEBUG -x NCCL_IB_DISABLE -x NCCL_P2P_DISABLE -x PATH -x PYTHONPATH -x TENSORRT_LLM_CACHE_DIR -x TOKENIZERS_PARALLELISM -x TRANSFORMERS_CACHE -x TRTLLM_USE_UCX_KVCACHE -x USER bash -c 'trtllm-llmapi-launch serve --model test'",
}, },
} }
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment