Skip to content

Fix UCCL's consistency checks - #1151

Merged
brminich merged 10 commits into
ai-dynamo:mainfrom
praveingk:fix_uccl_consistency
Jan 15, 2026
Merged

brminich merged 10 commits into
ai-dynamo:mainfrom
praveingk:fix_uccl_consistency

Conversation

@praveingk

Copy link
Copy Markdown
Contributor

What?

Fix consistency check of UCCL

How?

UCCL backend was using the base address instead of the specific local/remote address during batches. This PR fixes that address and keeps multiple FIFO items for the batches instead of the single one used per handle.
A fix was also added to UCCL to enable memory check based on its range.

Signed-off-by: Pravein Govindan Kannan <pravein.govindan.kannan@ibm.com>
Signed-off-by: Pravein Govindan Kannan <pravein.govindan.kannan@ibm.com>
Signed-off-by: Pravein Govindan Kannan <pravein.govindan.kannan@ibm.com>
@praveingk
praveingk requested a review from a team as a code owner December 23, 2025 05:21
@copy-pr-bot

copy-pr-bot Bot commented Dec 23, 2025

Copy link
Copy Markdown

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

@github-actions

Copy link
Copy Markdown

👋 Hi praveingk! Thank you for contributing to ai-dynamo/nixl.

Your PR reviewers will review your contribution then trigger the CI to test your changes.

🚀

@brminich

Copy link
Copy Markdown
Contributor

/ok to test 9f0ab11

@brminich

Copy link
Copy Markdown
Contributor

/build

Comment thread src/plugins/uccl/uccl_backend.cpp Outdated
Comment thread src/plugins/uccl/uccl_backend.cpp Outdated
Comment thread src/plugins/uccl/uccl_backend.cpp Outdated
Comment thread src/plugins/uccl/uccl_backend.cpp Outdated
@brminich

Copy link
Copy Markdown
Contributor

i see this failure, which may be related. Checking if the CI node is healthy

[2025-12-23T08:30:49.380Z] + run_nixlbench --benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency
[2025-12-23T08:30:49.380Z] + args='--benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency'
[2025-12-23T08:30:49.380Z] + ./bin/nixlbench --etcd-endpoints http://127.0.0.1:12510 --filepath /tmp --total_buffer_size 80000000 --start_block_size 4096 --max_block_size 16384 --start_batch_size 1 --max_batch_size 4 --benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency
[2025-12-23T08:30:49.380Z] WARNING: Adjusting num_iter to 1008 to allow equal distribution to 1 threads
[2025-12-23T08:30:49.380Z] WARNING: Adjusting warmup_iter to 112 to allow equal distribution to 1 threads
[2025-12-23T08:30:49.380Z] Connecting to ETCD at http://127.0.0.1:12510
[2025-12-23T08:30:49.380Z] ETCD Runtime: Registered as rank 0 item 1 of 2
[2025-12-23T08:30:49.380Z] UCCL backend
[2025-12-23T08:30:49.380Z] Creating Engine with CPUs: 4
[2025-12-23T08:30:49.646Z] [WARN] safe_pcie_distance failed: filesystem error: cannot make canonical path: Invalid argument []
[2025-12-23T08:30:49.646Z] RdmaDeviceManager: Found 1 RDMA device(s)
[2025-12-23T08:30:49.646Z]   [0] mlx5_0
[2025-12-23T08:30:49.646Z] RdmaDeviceManager: Initialization complete
[2025-12-23T08:30:49.646Z] nixlbench: efa/rdma_device.h:93: std::vector<long unsigned int> RdmaDeviceManager::get_best_dev_idx(int): Assertion `candidates.size() == 4' failed.
[2025-12-23T08:30:49.646Z] *** Aborted at 1766478649 (unix time) try "date -d @1766478649" if you are using GNU date ***
[2025-12-23T08:30:49.646Z] PC: @                0x0 (unknown)
[2025-12-23T08:30:49.646Z] *** SIGABRT (@0x242650002446e) received by PID 148590 (TID 0x724906831000) from PID 148590; stack trace: ***
[2025-12-23T08:30:49.646Z]     @     0x7248bc7e1206 (unknown)
[2025-12-23T08:30:49.646Z]     @     0x724907f6c330 (unknown)
[2025-12-23T08:30:49.646Z]     @     0x724907fc5b2c pthread_kill
[2025-12-23T08:30:49.646Z]     @     0x724907f6c27e gsignal
[2025-12-23T08:30:49.646Z]     @     0x724907f4f8ff abort
[2025-12-23T08:30:49.646Z]     @     0x724907f4f81b (unknown)
[2025-12-23T08:30:49.646Z]     @     0x724907f62517 __assert_fail
[2025-12-23T08:30:49.646Z]     @     0x72487d2b08f5 RdmaDeviceManager::get_best_dev_idx()
[2025-12-23T08:30:49.646Z]     @     0x72487d2b136f EFAEndpoint::EFAEndpoint()
[2025-12-23T08:30:49.646Z]     @     0x72487d278806 Endpoint::Endpoint()
[2025-12-23T08:30:49.646Z]     @     0x72487d242003 uccl_engine_create()
[2025-12-23T08:30:49.646Z]     @     0x72487d3704aa nixlUcclEngine::nixlUcclEngine()
[2025-12-23T08:30:49.646Z]     @     0x72487d388ec5 nixlBackendPluginCreator<>::createEngine()
[2025-12-23T08:30:49.646Z]     @     0x72490ea34627 nixlBackendPluginHandle::createEngine()
[2025-12-23T08:30:49.646Z]     @     0x72490e9fffb1 nixlAgent::createBackend()
[2025-12-23T08:30:49.646Z]     @     0x57ca0f184a41 xferBenchNixlWorker::xferBenchNixlWorker()
[2025-12-23T08:30:49.646Z]     @     0x57ca0f167747 std::make_unique<>()
[2025-12-23T08:30:49.646Z]     @     0x57ca0f16390a main
[2025-12-23T08:30:49.646Z]     @     0x724907f511ca (unknown)
[2025-12-23T08:30:49.646Z]     @     0x724907f5128b __libc_start_main
[2025-12-23T08:30:49.646Z]     @     0x57ca0f165675 _start
[2025-12-23T08:30:49.646Z] .gitlab/test_nixlbench.sh: line 64: 148590 Aborted                 (core dumped) ./bin/nixlbench --etcd-endpoints ${NIXL_ETCD_ENDPOINTS} $DEFAULT_NB_PARAMS $args
[2025-12-23T08:30:50.227Z] + run_nixlbench --benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency
[2025-12-23T08:30:50.227Z] + args='--benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency'
[2025-12-23T08:30:50.227Z] + ./bin/nixlbench --etcd-endpoints http://127.0.0.1:12510 --filepath /tmp --total_buffer_size 80000000 --start_block_size 4096 --max_block_size 16384 --start_batch_size 1 --max_batch_size 4 --benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency
[2025-12-23T08:30:50.227Z] WARNING: Adjusting num_iter to 1008 to allow equal distribution to 1 threads
[2025-12-23T08:30:50.227Z] WARNING: Adjusting warmup_iter to 112 to allow equal distribution to 1 threads
[2025-12-23T08:30:50.227Z] Connecting to ETCD at http://127.0.0.1:12510
[2025-12-23T08:30:50.227Z] ETCD Runtime: Registered as rank 1 item 2 of 2
[2025-12-23T08:30:50.493Z] UCCL backend
[2025-12-23T08:30:50.493Z] Creating Engine with CPUs: 4
[2025-12-23T08:30:50.493Z] RdmaDeviceManager: Found 1 RDMA device(s)
[2025-12-23T08:30:50.493Z]   [0] mlx5_0
[2025-12-23T08:30:50.493Z] RdmaDeviceManager: Initialization complete
[2025-12-23T08:30:50.493Z] [WARN] safe_pcie_distance failed: filesystem error: cannot make canonical path: Invalid argument []
[2025-12-23T08:30:50.493Z] nixlbench: efa/rdma_device.h:93: std::vector<long unsigned int> RdmaDeviceManager::get_best_dev_idx(int): Assertion `candidates.size() == 4' failed.
[2025-12-23T08:30:50.493Z] *** Aborted at 1766478650 (unix time) try "date -d @1766478650" if you are using GNU date ***
[2025-12-23T08:30:50.493Z] PC: @                0x0 (unknown)
[2025-12-23T08:30:50.493Z] *** SIGABRT (@0x2426500024498) received by PID 148632 (TID 0x7cb543285000) from PID 148632; stack trace: ***
[2025-12-23T08:30:50.493Z]     @     0x7cb504264206 (unknown)
[2025-12-23T08:30:50.493Z]     @     0x7cb544b16330 (unknown)
[2025-12-23T08:30:50.493Z]     @     0x7cb544b6fb2c pthread_kill
[2025-12-23T08:30:50.493Z]     @     0x7cb544b1627e gsignal
[2025-12-23T08:30:50.493Z]     @     0x7cb544af98ff abort
[2025-12-23T08:30:50.493Z]     @     0x7cb544af981b (unknown)
[2025-12-23T08:30:50.493Z]     @     0x7cb544b0c517 __assert_fail
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1d048f5 RdmaDeviceManager::get_best_dev_idx()
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1d0536f EFAEndpoint::EFAEndpoint()
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1ccc806 Endpoint::Endpoint()
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1c96003 uccl_engine_create()
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1dc44aa nixlUcclEngine::nixlUcclEngine()
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1ddcec5 nixlBackendPluginCreator<>::createEngine()
[2025-12-23T08:30:50.493Z]     @     0x7cb54b488627 nixlBackendPluginHandle::createEngine()
[2025-12-23T08:30:50.493Z]     @     0x7cb54b453fb1 nixlAgent::createBackend()
[2025-12-23T08:30:50.493Z]     @     0x5e0d8fc7ea41 xferBenchNixlWorker::xferBenchNixlWorker()
[2025-12-23T08:30:50.493Z]     @     0x5e0d8fc61747 std::make_unique<>()
[2025-12-23T08:30:50.493Z]     @     0x5e0d8fc5d90a main
[2025-12-23T08:30:50.493Z]     @     0x7cb544afb1ca (unknown)
[2025-12-23T08:30:50.493Z]     @     0x7cb544afb28b __libc_start_main
[2025-12-23T08:30:50.493Z]     @     0x5e0d8fc5f675 _start
[2025-12-23T08:30:50.760Z] .gitlab/test_nixlbench.sh: line 64: 148632 Aborted                 (core dumped) ./bin/nixlbench --etcd-endpoints ${NIXL_ETCD_ENDPOINTS} $DEFAULT_NB_PARAMS $args
[2025-12-23T08:30:53.117Z] + docker rm -f nixl-ci-gpu-101-2
[2025-12-23T08:31:01.316Z] nixl-ci-gpu-101-2
[2025-12-23T08:31:01.316Z] + docker image rm -f nixl-ci-gpu-101-2
[2025-12-23T08:31:01.316Z] Untagged: nixl-ci-gpu-101-2:latest
[2025-12-23T08:31:01.316Z] Deleted: sha256:e44d7bb179d4bf9c4b7cf9a1a2887e9d29cece8060fe66ba8780d5c6ead8b199
Test Nixlbench failed with msg: Step Test Nixlbench failed with exit code=134hudson.AbortException: parallel task failed with msg: hudson.AbortException: Test CPP failed with msg: Step Test CPP failed with exit code=1
Response Code :: 201
Response Code :: 204

@praveingk

Copy link
Copy Markdown
Contributor Author

i see this failure, which may be related. Checking if the CI node is healthy

[2025-12-23T08:30:49.380Z] + run_nixlbench --benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency
[2025-12-23T08:30:49.380Z] + args='--benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency'
[2025-12-23T08:30:49.380Z] + ./bin/nixlbench --etcd-endpoints http://127.0.0.1:12510 --filepath /tmp --total_buffer_size 80000000 --start_block_size 4096 --max_block_size 16384 --start_batch_size 1 --max_batch_size 4 --benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency
[2025-12-23T08:30:49.380Z] WARNING: Adjusting num_iter to 1008 to allow equal distribution to 1 threads
[2025-12-23T08:30:49.380Z] WARNING: Adjusting warmup_iter to 112 to allow equal distribution to 1 threads
[2025-12-23T08:30:49.380Z] Connecting to ETCD at http://127.0.0.1:12510
[2025-12-23T08:30:49.380Z] ETCD Runtime: Registered as rank 0 item 1 of 2
[2025-12-23T08:30:49.380Z] UCCL backend
[2025-12-23T08:30:49.380Z] Creating Engine with CPUs: 4
[2025-12-23T08:30:49.646Z] [WARN] safe_pcie_distance failed: filesystem error: cannot make canonical path: Invalid argument []
[2025-12-23T08:30:49.646Z] RdmaDeviceManager: Found 1 RDMA device(s)
[2025-12-23T08:30:49.646Z]   [0] mlx5_0
[2025-12-23T08:30:49.646Z] RdmaDeviceManager: Initialization complete
[2025-12-23T08:30:49.646Z] nixlbench: efa/rdma_device.h:93: std::vector<long unsigned int> RdmaDeviceManager::get_best_dev_idx(int): Assertion `candidates.size() == 4' failed.
[2025-12-23T08:30:49.646Z] *** Aborted at 1766478649 (unix time) try "date -d @1766478649" if you are using GNU date ***
[2025-12-23T08:30:49.646Z] PC: @                0x0 (unknown)
[2025-12-23T08:30:49.646Z] *** SIGABRT (@0x242650002446e) received by PID 148590 (TID 0x724906831000) from PID 148590; stack trace: ***
[2025-12-23T08:30:49.646Z]     @     0x7248bc7e1206 (unknown)
[2025-12-23T08:30:49.646Z]     @     0x724907f6c330 (unknown)
[2025-12-23T08:30:49.646Z]     @     0x724907fc5b2c pthread_kill
[2025-12-23T08:30:49.646Z]     @     0x724907f6c27e gsignal
[2025-12-23T08:30:49.646Z]     @     0x724907f4f8ff abort
[2025-12-23T08:30:49.646Z]     @     0x724907f4f81b (unknown)
[2025-12-23T08:30:49.646Z]     @     0x724907f62517 __assert_fail
[2025-12-23T08:30:49.646Z]     @     0x72487d2b08f5 RdmaDeviceManager::get_best_dev_idx()
[2025-12-23T08:30:49.646Z]     @     0x72487d2b136f EFAEndpoint::EFAEndpoint()
[2025-12-23T08:30:49.646Z]     @     0x72487d278806 Endpoint::Endpoint()
[2025-12-23T08:30:49.646Z]     @     0x72487d242003 uccl_engine_create()
[2025-12-23T08:30:49.646Z]     @     0x72487d3704aa nixlUcclEngine::nixlUcclEngine()
[2025-12-23T08:30:49.646Z]     @     0x72487d388ec5 nixlBackendPluginCreator<>::createEngine()
[2025-12-23T08:30:49.646Z]     @     0x72490ea34627 nixlBackendPluginHandle::createEngine()
[2025-12-23T08:30:49.646Z]     @     0x72490e9fffb1 nixlAgent::createBackend()
[2025-12-23T08:30:49.646Z]     @     0x57ca0f184a41 xferBenchNixlWorker::xferBenchNixlWorker()
[2025-12-23T08:30:49.646Z]     @     0x57ca0f167747 std::make_unique<>()
[2025-12-23T08:30:49.646Z]     @     0x57ca0f16390a main
[2025-12-23T08:30:49.646Z]     @     0x724907f511ca (unknown)
[2025-12-23T08:30:49.646Z]     @     0x724907f5128b __libc_start_main
[2025-12-23T08:30:49.646Z]     @     0x57ca0f165675 _start
[2025-12-23T08:30:49.646Z] .gitlab/test_nixlbench.sh: line 64: 148590 Aborted                 (core dumped) ./bin/nixlbench --etcd-endpoints ${NIXL_ETCD_ENDPOINTS} $DEFAULT_NB_PARAMS $args
[2025-12-23T08:30:50.227Z] + run_nixlbench --benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency
[2025-12-23T08:30:50.227Z] + args='--benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency'
[2025-12-23T08:30:50.227Z] + ./bin/nixlbench --etcd-endpoints http://127.0.0.1:12510 --filepath /tmp --total_buffer_size 80000000 --start_block_size 4096 --max_block_size 16384 --start_batch_size 1 --max_batch_size 4 --benchmark_group H2y6U74FTOpYj97wUmPg6mMlLy0m5LQN --backend UCCL --op_type READ --initiator_seg_type VRAM --target_seg_type VRAM --check_consistency
[2025-12-23T08:30:50.227Z] WARNING: Adjusting num_iter to 1008 to allow equal distribution to 1 threads
[2025-12-23T08:30:50.227Z] WARNING: Adjusting warmup_iter to 112 to allow equal distribution to 1 threads
[2025-12-23T08:30:50.227Z] Connecting to ETCD at http://127.0.0.1:12510
[2025-12-23T08:30:50.227Z] ETCD Runtime: Registered as rank 1 item 2 of 2
[2025-12-23T08:30:50.493Z] UCCL backend
[2025-12-23T08:30:50.493Z] Creating Engine with CPUs: 4
[2025-12-23T08:30:50.493Z] RdmaDeviceManager: Found 1 RDMA device(s)
[2025-12-23T08:30:50.493Z]   [0] mlx5_0
[2025-12-23T08:30:50.493Z] RdmaDeviceManager: Initialization complete
[2025-12-23T08:30:50.493Z] [WARN] safe_pcie_distance failed: filesystem error: cannot make canonical path: Invalid argument []
[2025-12-23T08:30:50.493Z] nixlbench: efa/rdma_device.h:93: std::vector<long unsigned int> RdmaDeviceManager::get_best_dev_idx(int): Assertion `candidates.size() == 4' failed.
[2025-12-23T08:30:50.493Z] *** Aborted at 1766478650 (unix time) try "date -d @1766478650" if you are using GNU date ***
[2025-12-23T08:30:50.493Z] PC: @                0x0 (unknown)
[2025-12-23T08:30:50.493Z] *** SIGABRT (@0x2426500024498) received by PID 148632 (TID 0x7cb543285000) from PID 148632; stack trace: ***
[2025-12-23T08:30:50.493Z]     @     0x7cb504264206 (unknown)
[2025-12-23T08:30:50.493Z]     @     0x7cb544b16330 (unknown)
[2025-12-23T08:30:50.493Z]     @     0x7cb544b6fb2c pthread_kill
[2025-12-23T08:30:50.493Z]     @     0x7cb544b1627e gsignal
[2025-12-23T08:30:50.493Z]     @     0x7cb544af98ff abort
[2025-12-23T08:30:50.493Z]     @     0x7cb544af981b (unknown)
[2025-12-23T08:30:50.493Z]     @     0x7cb544b0c517 __assert_fail
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1d048f5 RdmaDeviceManager::get_best_dev_idx()
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1d0536f EFAEndpoint::EFAEndpoint()
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1ccc806 Endpoint::Endpoint()
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1c96003 uccl_engine_create()
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1dc44aa nixlUcclEngine::nixlUcclEngine()
[2025-12-23T08:30:50.493Z]     @     0x7cb4b1ddcec5 nixlBackendPluginCreator<>::createEngine()
[2025-12-23T08:30:50.493Z]     @     0x7cb54b488627 nixlBackendPluginHandle::createEngine()
[2025-12-23T08:30:50.493Z]     @     0x7cb54b453fb1 nixlAgent::createBackend()
[2025-12-23T08:30:50.493Z]     @     0x5e0d8fc7ea41 xferBenchNixlWorker::xferBenchNixlWorker()
[2025-12-23T08:30:50.493Z]     @     0x5e0d8fc61747 std::make_unique<>()
[2025-12-23T08:30:50.493Z]     @     0x5e0d8fc5d90a main
[2025-12-23T08:30:50.493Z]     @     0x7cb544afb1ca (unknown)
[2025-12-23T08:30:50.493Z]     @     0x7cb544afb28b __libc_start_main
[2025-12-23T08:30:50.493Z]     @     0x5e0d8fc5f675 _start
[2025-12-23T08:30:50.760Z] .gitlab/test_nixlbench.sh: line 64: 148632 Aborted                 (core dumped) ./bin/nixlbench --etcd-endpoints ${NIXL_ETCD_ENDPOINTS} $DEFAULT_NB_PARAMS $args
[2025-12-23T08:30:53.117Z] + docker rm -f nixl-ci-gpu-101-2
[2025-12-23T08:31:01.316Z] nixl-ci-gpu-101-2
[2025-12-23T08:31:01.316Z] + docker image rm -f nixl-ci-gpu-101-2
[2025-12-23T08:31:01.316Z] Untagged: nixl-ci-gpu-101-2:latest
[2025-12-23T08:31:01.316Z] Deleted: sha256:e44d7bb179d4bf9c4b7cf9a1a2887e9d29cece8060fe66ba8780d5c6ead8b199
Test Nixlbench failed with msg: Step Test Nixlbench failed with exit code=134hudson.AbortException: parallel task failed with msg: hudson.AbortException: Test CPP failed with msg: Step Test CPP failed with exit code=1
Response Code :: 201
Response Code :: 204

By any chance is it an AWS EFA node? Wondering why EFAEndpoint is getting triggered.

Signed-off-by: Pravein Govindan Kannan <pravein.govindan.kannan@ibm.com>
Signed-off-by: Pravein Govindan Kannan <pravein.govindan.kannan@ibm.com>
@brminich

Copy link
Copy Markdown
Contributor

/ok to test 353baf0

@brminich

Copy link
Copy Markdown
Contributor

/build

@praveingk

Copy link
Copy Markdown
Contributor Author

@brminich Recently experimental EFA feature was pushed in UCCL. Probably, here it's hitting that on the AWS nodes. Can you please confirm if it's AWS nodes, and it was used previously? If so, I will need to turn off the EFA feature.

@brminich

Copy link
Copy Markdown
Contributor

@praveingk, no it is not AWS node. It is x86 with 1 h100 gpu and 1 BF3 NIC

@brminich

brminich commented Dec 23, 2025

Copy link
Copy Markdown
Contributor

can you open this link https://nbuprod.blsm.nvidia.com/nbu-swx-nixl-main/blue/rest/organizations/jenkins/pipelines/nixl-ci-gpu/runs/108/nodes/199/log/?start=0?

it contains some debug info including ibv_devinfo output

@brminich
brminich merged commit 151b6ab into ai-dynamo:main Jan 15, 2026
20 of 21 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants