Skip to content

Commit 0d8b7ae

Browse files
authored
[main][CI] update nightly VLLM_ENGINE_READY_TIMEOUT_S (vllm-project#11530)
### What this PR does / why we need it? This PR standardizes the `VLLM_ENGINE_READY_TIMEOUT_S` environment variable to `3000` across various nightly end-to-end (e2e) test configurations for DeepSeek and GLM models. This ensures consistent engine startup timeout limits across single-node and multi-node test environments. ### Does this PR introduce _any_ user-facing change? No, this change only affects internal nightly e2e test configurations. ### How was this patch tested? Tested via existing nightly e2e test suites. - vLLM version: v0.23.0 - vLLM main: vllm-project/vllm@1f486d9 Signed-off-by: pppeng <372907983@qq.com>
1 parent 0deb900 commit 0d8b7ae

7 files changed

Lines changed: 7 additions & 3 deletions

File tree

tests/e2e/nightly/multi_node/external_dp/config/GLM5_1-W8A8-EP-external.yaml

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -62,7 +62,7 @@ env_common: &env_common
6262
ACL_OP_INIT_MODE: "1"
6363
ASCEND_A3_ENABLE: "1"
6464
VLLM_NIXL_ABORT_REQUEST_TIMEOUT: "300000"
65-
VLLM_ENGINE_READY_TIMEOUT_S: "7200"
65+
VLLM_ENGINE_READY_TIMEOUT_S: "3000"
6666
HCCL_CONNECT_TIMEOUT: "1200"
6767
HCCL_INTRA_PCIE_ENABLE: "1"
6868
HCCL_INTRA_ROCE_ENABLE: "0"

tests/e2e/nightly/multi_node/internal_dp/config/DeepSeek-R1-W8A8-EPLB.yaml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -13,6 +13,7 @@ env_common: &env_common
1313
TASK_QUEUE_ENABLE: 1
1414
HCCL_OP_RETRY_ENABLE: "L0:0, L1:0, L2:0"
1515
DYNAMIC_EPLB: true
16+
VLLM_ENGINE_READY_TIMEOUT_S: 3000
1617

1718
disaggregated_prefill:
1819
enabled: true

tests/e2e/nightly/multi_node/internal_dp/config/DeepSeek-R1-W8A8-longseq.yaml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -14,6 +14,7 @@ env_common: &env_common
1414
TASK_QUEUE_ENABLE: 1
1515
HCCL_OP_RETRY_ENABLE: "L0:0, L1:0"
1616
VLLM_MOONCAKE_ABORT_REQUEST_TIMEOUT: 480
17+
VLLM_ENGINE_READY_TIMEOUT_S: 3000
1718

1819
disaggregated_prefill:
1920
enabled: true

tests/e2e/nightly/multi_node/internal_dp/config/DeepSeek-V3_2-W8A8-A3-dual-nodes.yaml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -13,6 +13,7 @@ env_common: &env_common
1313
VLLM_ASCEND_ENABLE_MLAPO: 1
1414
VLLM_ASCEND_ENABLE_FLASHCOMM1: 1
1515
ASCEND_A3_EBA_ENABLE: 1
16+
VLLM_ENGINE_READY_TIMEOUT_S: 3000
1617

1718
# TODO: need to identify why TP and mtp+1 divisibility rules break on dual-node case
1819

tests/e2e/nightly/multi_node/internal_dp/config/DeepSeek-V3_2-W8A8-EP.yaml

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -16,7 +16,7 @@ env_common: &env_common
1616
ACL_OP_INIT_MODE: 1
1717
ASCEND_A3_ENABLE: 1
1818
VLLM_MOONCAKE_ABORT_REQUEST_TIMEOUT: 480
19-
VLLM_ENGINE_READY_TIMEOUT_S: 1800
19+
VLLM_ENGINE_READY_TIMEOUT_S: 3000
2020
HCCL_CONNECT_TIMEOUT: 1200
2121

2222

tests/e2e/nightly/multi_node/internal_dp/config/GLM5_1-W8A8-EP.yaml

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -16,7 +16,7 @@ env_common: &env_common
1616
ACL_OP_INIT_MODE: 1
1717
ASCEND_A3_ENABLE: 1
1818
VLLM_NIXL_ABORT_REQUEST_TIMEOUT: "300000"
19-
VLLM_ENGINE_READY_TIMEOUT_S: "7200"
19+
VLLM_ENGINE_READY_TIMEOUT_S: "3000"
2020
HCCL_CONNECT_TIMEOUT: "1200"
2121
HCCL_INTRA_PCIE_ENABLE: 1
2222
HCCL_INTRA_ROCE_ENABLE: 0

tests/e2e/nightly/single_node/models/configs/DeepSeek-V3.2-W8A8.yaml

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -21,6 +21,7 @@ test_cases:
2121
VLLM_ASCEND_ENABLE_MLAPO: "1"
2222
VLLM_ASCEND_ENABLE_FLASHCOMM1: "1"
2323
SERVER_PORT: "DEFAULT_PORT"
24+
VLLM_ENGINE_READY_TIMEOUT_S: "3000"
2425
server_cmd:
2526
- "--tensor-parallel-size"
2627
- "8"

0 commit comments

Comments
 (0)