[GH-ISSUE #202] [ERROR] Run ``test_low_latency.py`` failed in both one node and two nodes #165

New issue

Closed

opened 2026-10-26 09:55:30 +00:00 by bsakaguchi · 2 comments

bsakaguchi commented 2026-10-26 09:55:30 +00:00

Owner

Copy link

Originally created by @Tyndalllll on GitHub (Jun 10, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/202

Description

I'm running into a problem where I can't run test_internode.py and test_low_latency.py properly as well (the error for both runs is the same), whether it's one node or two nodes(H20 * 8). But I can run test_intranode.py normally.

Could you please help me with this? thanks for your excellent job.

Environment

Docker container from https://github.com/sgl-project/sglang/blob/main/docker/Dockerfile.deepep

nvidia-smi topo -mp

        GPU0    GPU1    GPU2    GPU3    GPU4    GPU5    GPU6    GPU7    CPU Affinity    NUMA Affinity   GPU NUMA ID
GPU0     X      NODE    NODE    NODE    SYS     SYS     SYS     SYS     0-47,96-143     0               N/A
GPU1    NODE     X      NODE    NODE    SYS     SYS     SYS     SYS     0-47,96-143     0               N/A
GPU2    NODE    NODE     X      NODE    SYS     SYS     SYS     SYS     0-47,96-143     0               N/A
GPU3    NODE    NODE    NODE     X      SYS     SYS     SYS     SYS     0-47,96-143     0               N/A
GPU4    SYS     SYS     SYS     SYS      X      NODE    NODE    NODE    48-95,144-191   1               N/A
GPU5    SYS     SYS     SYS     SYS     NODE     X      NODE    NODE    48-95,144-191   1               N/A
GPU6    SYS     SYS     SYS     SYS     NODE    NODE     X      NODE    48-95,144-191   1               N/A
GPU7    SYS     SYS     SYS     SYS     NODE    NODE    NODE     X      48-95,144-191   1               N/A

Legend:

  X    = Self
  SYS  = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI)
  NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node
  PHB  = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU)
  PXB  = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge)
  PIX  = Connection traversing at most a single PCIe bridge

ibv_devinfo

hca_id: erdma_0
        transport:                      eRDMA (0)
        fw_ver:                         0.2.0
        node_guid:                      0216:3eff:fe43:24be
        sys_image_guid:                 0216:3eff:fe43:24be
        vendor_id:                      0x1ded
        vendor_part_id:                 4223
        hw_ver:                         0x0
        phys_port_cnt:                  1
                port:   1
                        state:                  PORT_ACTIVE (4)
                        max_mtu:                4096 (5)
                        active_mtu:             4096 (5)
                        sm_lid:                 0
                        port_lid:               0
                        port_lmc:               0x00
                        link_layer:             Ethernet

gdrcopy_copybw

GPU id:0; name: NVIDIA H20; Bus id: 0000:08:00
GPU id:1; name: NVIDIA H20; Bus id: 0000:7e:00
GPU id:2; name: NVIDIA H20; Bus id: 0000:a2:00
GPU id:3; name: NVIDIA H20; Bus id: 0000:c6:00
GPU id:4; name: NVIDIA H20; Bus id: 0001:09:00
GPU id:5; name: NVIDIA H20; Bus id: 0001:7f:00
GPU id:6; name: NVIDIA H20; Bus id: 0001:a3:00
GPU id:7; name: NVIDIA H20; Bus id: 0001:c7:00
selecting device 0
testing size: 131072
rounded size: 131072
gpu alloc fn: cuMemAlloc
device ptr: 7f15a7400000
map_d_ptr: 0x7f17c8238000
info.va: 7f15a7400000
info.mapped_size: 131072
info.page_size: 65536
info.mapped: 1
info.wc_mapping: 1
page offset: 0
user-space pointer:0x7f17c8238000
writing test, size=131072 offset=0 num_iters=10000
write BW: 17649.5MB/s
reading test, size=131072 offset=0 num_iters=100
read BW: 712.677MB/s
unmapping buffer
unpinning buffer
closing gdrdrv

nvshmem-info -a

NVSHMEM v3.2.5

Build Information:
  CUDA API                     12040
  CUDA Driver                  12080
  Build Timestamp              Jun  9 2025 09:20:22
  Build Variables             
        NVSHMEM_DEBUG=OFF NVSHMEM_DEVEL=OFF NVSHMEM_DEFAULT_PMI2=OFF
        NVSHMEM_DEFAULT_PMIX=OFF NVSHMEM_DEFAULT_UCX=OFF
        NVSHMEM_ENABLE_ALL_DEVICE_INLINING=OFF NVSHMEM_GPU_COLL_USE_LDST=OFF
        NVSHMEM_IBGDA_SUPPORT=ON NVSHMEM_IBGDA_SUPPORT_GPUMEM_ONLY=OFF
        NVSHMEM_IBDEVX_SUPPORT=OFF NVSHMEM_IBRC_SUPPORT=ON
        NVSHMEM_MPI_SUPPORT=OFF NVSHMEM_NVTX=ON NVSHMEM_PMIX_SUPPORT=OFF
        NVSHMEM_SHMEM_SUPPORT=OFF NVSHMEM_TEST_STATIC_LIB=OFF
        NVSHMEM_TIMEOUT_DEVICE_POLLING=OFF NVSHMEM_TRACE=OFF NVSHMEM_UCX_SUPPORT=OFF
        NVSHMEM_USE_DLMALLOC=OFF NVSHMEM_USE_NCCL=OFF NVSHMEM_USE_GDRCOPY=ON
        NVSHMEM_VERBOSE=OFF CUDA_HOME=/usr/local/cuda
        GDRCOPY_HOME=/sgl-workspace/gdrcopy-2.5 LIBFABRIC_HOME=/usr/local/libfabric
        MPI_HOME=/usr/local/ompi NCCL_HOME=/usr/local/nccl
        NVSHMEM_PREFIX=/usr/local/nvshmem PMIX_HOME=/usr SHMEM_HOME=/usr/local/ompi
        UCX_HOME=/usr/local/ucx

Standard options:
  NVSHMEM_VERSION              false (type: bool, default: false)
        Print library version at startup
  NVSHMEM_INFO                 false (type: bool, default: false)
        Print environment variable options at startup
  NVSHMEM_DISABLE_NVLS         false (type: bool, default: false)
        Disable NVLS SHARP resources for collectives, even if available for platform
  NVSHMEM_SYMMETRIC_SIZE       1073741824 (type: size, default: 1073741824)
        Specifies the size (in bytes) of the symmetric heap memory per PE. The
        size is implementation-defined and must be at least as large as the integer
        ceiling of the product of the numeric prefix and the scaling factor. The
        character suffixes for the scaling factor are as follows:

          *  k or K multiplies by 2^10 (kibibytes)
          *  m or M multiplies by 2^20 (mebibytes)
          *  g or G multiplies by 2^30 (gibibytes)
          *  t or T multiplies by 2^40 (tebibytes)

        For example, string '20m' is equivalent to the integer value 20971520, or 20
        mebibytes. Similarly the string '3.1M' is equivalent to the integer value
        3250586. Only one multiplier is recognized and any characters following the
        multiplier are ignored, so '20kk' will not produce the same result as '20m'.
        Usage of string '.5m' will yield the same result as the string '0.5m'.
        An invalid value for NVSHMEM_SYMMETRIC_SIZE is an error, which the NVSHMEM
        library shall report by either returning a nonzero value from
        nvshmem_init_thread or causing program termination.
  NVSHMEM_DEBUG                "" (type: string, default: "")
        Set to enable debugging messages.
        Optional values: VERSION, WARN, INFO, ABORT, TRACE

Bootstrap options:
  NVSHMEM_BOOTSTRAP            "PMI" (type: string, default: "PMI")
        Name of the default bootstrap that should be used to initialize NVSHMEM.
        Allowed values: PMI, MPI, SHMEM, plugin, UID
  NVSHMEM_BOOTSTRAP_PMI        "PMI" (type: string, default: "PMI")
        Name of the PMI bootstrap that should be used to initialize NVSHMEM.
        Allowed values: PMI, PMI-2, PMIX
  NVSHMEM_BOOTSTRAP_PLUGIN     "" (type: string, default: "")
        Absolute path to or name of the bootstrap plugin file to load when
        NVSHMEM_BOOTSTRAP=plugin is specified
  NVSHMEM_BOOTSTRAP_MPI_PLUGIN "nvshmem_bootstrap_mpi.so.3" (type: string, default: "nvshmem_bootstrap_mpi.so.3")
        Absolute path to or name of the MPI bootstrap plugin file. 
        NVSHMEM will search for the plugin based on linux linker priorities. See man
        dlopen
  NVSHMEM_BOOTSTRAP_SHMEM_PLUGIN "nvshmem_bootstrap_shmem.so.3" (type: string, default: "nvshmem_bootstrap_shmem.so.3")
        Absolute path to or name of the SHMEM bootstrap plugin file. 
        NVSHMEM will search for the plugin based on linux linker priorities. See man
        dlopen
  NVSHMEM_BOOTSTRAP_PMI_PLUGIN "nvshmem_bootstrap_pmi.so.3" (type: string, default: "nvshmem_bootstrap_pmi.so.3")
        Absolute path to or name of the PMI bootstrap plugin file. 
        NVSHMEM will search for the plugin based on linux linker priorities. See man
        dlopen
  NVSHMEM_BOOTSTRAP_PMI2_PLUGIN "nvshmem_bootstrap_pmi2.so.3" (type: string, default: "nvshmem_bootstrap_pmi2.so.3")
        Absolute path to or name of the PMI-2 bootstrap plugin file. 
        NVSHMEM will search for the plugin based on linux linker priorities. See man
        dlopen
  NVSHMEM_BOOTSTRAP_PMIX_PLUGIN "nvshmem_bootstrap_pmix.so.3" (type: string, default: "nvshmem_bootstrap_pmix.so.3")
        Absolute path to or name of the PMIx bootstrap plugin file. 
        NVSHMEM will search for the plugin based on linux linker priorities. See man
        dlopen
  NVSHMEM_BOOTSTRAP_UID_PLUGIN "nvshmem_bootstrap_uid.so.3" (type: string, default: "nvshmem_bootstrap_uid.so.3")
        Absolute path to or name of the UID bootstrap plugin file. 
        NVSHMEM will search for the plugin based on linux linker priorities. See man
        dlopen

Additional options:
  NVSHMEM_CUDA_PATH            "" (type: string, default: "")
        Path to directory containing libcuda.so (for use when not in default location)
  NVSHMEM_DEBUG_ATTACH_DELAY   0 (type: int, default: 0)
        Delay (in seconds) during the first call to NVSHMEM_INIT to allow for attaching
        a debuggger (Default 0)
  NVSHMEM_DEBUG_FILE           "" (type: string, default: "")
        Debugging output filename, may contain %h for hostname and %p for pid
  NVSHMEM_MAX_TEAMS            32 (type: long, default: 32)
        Maximum number of simultaneous teams allowed
  NVSHMEM_MAX_MEMORY_PER_GPU   137438953472 (type: size, default: 137438953472)
        Maximum memory per GPU
  NVSHMEM_DISABLE_CUDA_VMM     false (type: bool, default: false)
        Disable use of CUDA VMM for P2P memory mapping. By default, CUDA VMM is enabled
        on x86 and disabled on P9. CUDA VMM feature in NVSHMEM requires CUDA RT version
        and CUDA Driver version to be greater than or equal to 11.3.
  NVSHMEM_DISABLE_P2P          false (type: bool, default: false)
        Disable P2P connectivity of GPUs even when available
  NVSHMEM_IGNORE_CUDA_MPS_ACTIVE_THREAD_PERCENTAGE false (type: bool, default: false)
        When doing Multi-Process Per GPU (MPG) run, full API support is available only
        if sum of CUDA_MPS_ACTIVE_THREAD_PERCENTAGE of processes running on a GPU is <=
        100%. Through this variable, user can request NVSHMEM runtime to ignore the
        active thread percentage and allow full MPG support. Users enable it at their
        own risk as NVSHMEM might deadlock.
  NVSHMEM_CUMEM_GRANULARITY    536870912 (type: size, default: 536870912)
        Granularity for cuMemAlloc/cuMemCreate
  NVSHMEM_PROXY_REQUEST_BATCH_MAX 32 (type: int, default: 32)
        Maxmum number of requests that the proxy thread processes in a single iteration
        of the progress loop.

Collectives options:
  NVSHMEM_DISABLE_NCCL         false (type: bool, default: false)
        Disable use of NCCL for collective operations
  NVSHMEM_BARRIER_DISSEM_KVAL  2 (type: int, default: 2)
        Radix of the dissemination algorithm used for barriers
  NVSHMEM_BARRIER_TG_DISSEM_KVAL 2 (type: int, default: 2)
        Radix of the dissemination algorithm used for thread group barriers
  NVSHMEM_FCOLLECT_LL_THRESHOLD 2048 (type: size, default: 2048)
        Message size threshold up to which fcollect LL algo will be used

  NVSHMEM_REDUCE_SCRATCH_SIZE  524288 (type: size, default: 524288)
        Amount of symmetric heap memory (minimum 16B, multiple of 8B) reserved by
        runtime for every team to implement reduce and reducescatter collectives

  NVSHMEM_BCAST_ALGO           0 (type: int, default: 0)
        Broadcast algorithm to be used.
          * 0 - use default algorithm selection strategy

  NVSHMEM_REDMAXLOC_ALGO       1 (type: int, default: 1)
        Reduction algorithm to be used for MAXLOC operation.
          * 1 - default, flag alltoall algorithm
          * 2 - flat reduce + flat bcast
          * 3 - topo-aware two-level reduce + topo-aware bcast


Transport options:
  NVSHMEM_REMOTE_TRANSPORT     "ibrc" (type: string, default: "ibrc")
        Selected transport for remote operations: ibrc, ucx, libfabric, ibdevx, none
  NVSHMEM_ENABLE_NIC_PE_MAPPING false (type: bool, default: false)
        When not set or set to 0, a PE is assigned the NIC on the node that is closest
        to it by distance. When set to 1, NVSHMEM either assigns NICs to PEs on a
        round-robin basis or uses NVSHMEM_HCA_PE_MAPPING or NVSHMEM_HCA_LIST when they
        are specified.
  NVSHMEM_DISABLE_LOCAL_ONLY_PROXY false (type: bool, default: false)
        When running on an NVLink-only configuaration (No-IB, No-UCX), completely
        disable the proxy thread. This will disable device side global exit and device
        side wait timeout polling (enabled by NVSHMEM_TIMEOUT_DEVICE_POLLING build-time
        variable) because these are processed by the proxy thread.
  NVSHMEM_IB_ENABLE_IBGDA      false (type: bool, default: false)
        Set to enable GPU-initiated communication transport.

Hidden options:
  NVSHMEM_INFO_HIDDEN          true (type: bool, default: false)
        Print hidden environment variable options at startup
  NVSHMEM_DISABLE_NVLS_SHARING true (type: bool, default: true)
        Disable NVLS SHARP resource sharing for user-defined teams
  NVSHMEM_HEAP_KIND            "DEVICE" (type: string, default: "DEVICE")
        Specify the memory kind used by the NVSHMEM symmetric heap.
        Allowed values: VIDMEM, SYSMEM
  NVSHMEM_ENABLE_RAIL_OPT      false (type: bool, default: false)
        Enable Rail Optimization when heap is in SYSMEM
  NVSHMEM_BOOTSTRAP_TWO_STAGE  false (type: bool, default: false)
        Ignore CUDA device setting during initialization,forcing two-stage
        initialization
  NVSHMEM_DEBUG_SUBSYS         "" (type: string, default: "")
        Comma separated list of debugging message sources. Prefix with '^' to exclude.
        Values: INIT, COLL, P2P, PROXY, TRANSPORT, MEM, BOOTSTRAP, TOPO, UTIL, ALL
  NVSHMEM_ENABLE_ERROR_CHECKS  false (type: bool, default: false)
        Enable error checks
  NVSHMEM_DISABLE_MNNVL        false (type: bool, default: false)
        Disable MNNVL connectivity for GPUs even when available
  NVSHMEM_CUMEM_HANDLE_TYPE    "FILE_DESCRIPTOR" (type: string, default: "FILE_DESCRIPTOR")
        Handle type for cuMemCreate. Supported are - FABRIC or FILE_DESCRIPTOR
  NVSHMEM_BYPASS_ACCESSIBILITY_CHECK false (type: bool, default: false)
        Bypass peer GPU accessbility checks
  NVSHMEM_FCOLLECT_NTHREADS    512 (type: int, default: 512)
        Sets number of threads per block for fcollect collective.
        By default, if no env is set, default value is min(max_occupancy per CTA, msg
        size per PE).
        If env is specified, value overrides the default irrespective of max occupancy
        per CTA

  NVSHMEM_REDUCESCATTER_NTHREADS 512 (type: int, default: 512)
        Sets number of threads per block for reducescatter collective.
        By default, if no env is set, default value is min(max_occupancy per CTA, msg
        size per PE).
        If env is specified, value overrides the default irrespective of max occupancy
        per CTA

  NVSHMEM_MAX_CTAS             1 (type: int, default: 1)
        Sets number of blocks per grid for host onstream collective.
        By default, if no env is set, default value to 1 CTA
        If env is specified, value overrides the default value

  NVSHMEM_REDUCE_RECEXCH_KVAL  2 (type: int, default: 2)
        Radix of the recursive exchange reduction algorithm
  NVSHMEM_FCOLLECT_LL128_THRESHOLD 0 (type: size, default: 0)
        Message size threshold up to which the fcollect LL128 algo will be used.
        LL128 will be used only when FCOLLECT_LL_THRESHOLD < size
  NVSHMEM_FCOLLECT_NVLS_THRESHOLD 16777216 (type: size, default: 16777216)
        Message size threshold up to which fcollect NVLS algo will be used

  NVSHMEM_REDUCESCATTER_NVLS_THRESHOLD 16777216 (type: size, default: 16777216)
        Message size threshold up to which reducescatter NVLS algo will be used

  NVSHMEM_BCAST_TREE_KVAL      2 (type: int, default: 2)
        Radix of the broadcast tree algorithm
  NVSHMEM_FCOLLECT_ALGO        0 (type: int, default: 0)
        Fcollect algorithm to be used. 
          * 0 - use default algorithm selection strategy

  NVSHMEM_REDUCE_ALGO          0 (type: int, default: 0)
        Allreduce algorithm to be used. 
           * 0/1 - use default algorithm selection strategy

  NVSHMEM_REDUCE_NVLS_THRESHOLD 2048 (type: int, default: 2048)
        Message size threshold up to which allreduce one-shot algo will be used

  NVSHMEM_REDUCESCATTER_ALGO   0 (type: int, default: 0)
        Reduce Scatter algorithm to be used. 
          * 0 - use default algorithm selection strategy

  NVSHMEM_ASSERT_ATOMICS_SYNC  false (type: bool, default: false)
        Bypass flush on wait_until at target
  NVSHMEM_BYPASS_FLUSH         false (type: bool, default: false)
        Bypass flush in proxy when enforcing consistency

NVTX options:
  NVSHMEM_NVTX                 "off" (type: string, default: "off")
        Set to enable NVTX instrumentation. Accepts a comma separated list of
        instrumentation groups. By default the NVTX instrumentation is disabled.
          init                : library setup
          alloc               : memory management
          launch              : kernel launch routines
          coll                : collective communications
          wait                : blocking point-to-point synchronization
          wait_on_stream      : point-to-point synchronization (on stream)
          test                : non-blocking point-to-point synchronization
          memorder            : memory ordering (quiet, fence)
          quiet_on_stream     : nvshmemx_quiet_on_stream
          atomic_fetch        : fetching atomic memory operations
          atomic_set          : non-fetchong atomic memory operations
          rma_blocking        : blocking remote memory access operations
          rma_nonblocking     : non-blocking remote memory access operations
          proxy               : activity of the proxy thread
          common              : init,alloc,launch,coll,memorder,wait,atomic_fetch,rma_blocking,proxy
          all                 : all groups
          off                 : disable all NVTX instrumentation

The error is as follows

1 node:

root@zz:/sgl-workspace/DeepEP# python3 tests/test_low_latency.py 
Allocating buffer size: 2115.111296 MB ...
WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping...

/sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA
WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping...

/sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA
WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping...

/sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA
WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping...

/sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA
WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping...

/sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA
WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping...

/sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA
WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping...

/sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA
WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping...

/sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA
/sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value srq creation failed 
srq creation failed 
srq creation failed 
srq creation failed 




/sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:1395: non-zero status: 7 /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:1395: non-zero status: 7 /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:1395: non-zero status: 7 /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:1395: non-zero status: 7 ep_create failed
ep_create failed
ep_create failed
ep_create failed




[iZ2ze1lasox0u2bv6k3iicZ:2136 :0:2136] Caught signal 11 (Segmentation fault: address not mapped to object at address 0x38)
[iZ2ze1lasox0u2bv6k3iicZ:2137 :0:2137] Caught signal 11 (Segmentation fault: address not mapped to object at address 0x38)
[iZ2ze1lasox0u2bv6k3iicZ:2142 :0:2142] Caught signal 11 (Segmentation fault: address not mapped to object at address 0x38)
[iZ2ze1lasox0u2bv6k3iicZ:2135 :0:2135] Caught signal 11 (Segmentation fault: address not mapped to object at address 0x38)
/sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value srq creation failed 
srq creation failed 
srq creation failed 
srq creation failed 

When I add NVSHMEM_DEBUG=1, it occurs some info (the log is too long so I show the abnormal ones):

/sgl-workspace/nvshmem/src/modules/transport/common/transport_gdr_common.cpp 73 GDR driver version: (2, 5)
/sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp 1632 Begin - Enumerating IB devices in the system ([<dev_id, device_name, num_ports>]) - 
/sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp 1652 Enumerated IB devices in the system - device id=0 (of 1), name=erdma_0, num_ports=1
/sgl-workspace/nvshmem/src/modules/transport/common/transport_ib_common.h 103 NVSHMEM_IB_ADDR_FAMILY set by environment to AF_INET
/sgl-workspace/nvshmem/src/modules/transport/common/transport_ib_common.h 126 NVSHMEM_IB_ADDR_RANGE set by environment to ::/0
/sgl-workspace/nvshmem/src/modules/transport/common/transport_ib_common.h 138 NET/IB: Ip address '::' is invalid for family AF_INET, ignoring address
/sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp 1739 End - Enumerating IB devices in the system

/sgl-workspace/nvshmem/src/modules/transport/ibgda/ibgda.cpp 3885 no active IB device that supports GPU-initiated communication is found, exiting...

2 nodes:

There is none output and hang until over time.

root@zz:/sgl-workspace/DeepEP#MASTER_ADDR= MASTER_ADDR WORLD_SIZE=2 RANK=0 python3 tests/test_internode.py 

When I use py-spy to output the stacks of all threads, it shows the thread hangs in init buffer of deepep:

root@zz:py-spy --dump -p 7422
Python v3.10.12 (/usr/bin/python3.10)

Thread 7422 (active): "MainThread"
    __torch_function__ (torch/utils/_device.py:104)
    all_gather_object (torch/distributed/distributed_c10d.py:3036)
    wrapper (torch/distributed/c10d_logger.py:81)
    __init__ (deep_ep/buffer.py:66)
    test_loop (test_internode.py:229)
    _wrap (torch/multiprocessing/spawn.py:90)
    run (multiprocessing/process.py:108)
    _bootstrap (multiprocessing/process.py:314)
    _main (multiprocessing/spawn.py:129)
    spawn_main (multiprocessing/spawn.py:116)
    <module> (<string>:1)

These problems originated from my use of sglang PD disggeraton to run deepseek. I found that enabling deepep would result in the inability to use multiple P nodes or D nodes, and the inability to enable the low_latency mode of the D node. For now, Deepep’s problem cannot be solved.

Is there anything reason?

Originally created by @Tyndalllll on GitHub (Jun 10, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/202 ### Description I'm running into a problem where I can't run ```test_internode.py``` and ```test_low_latency.py``` properly as well (**the error for both runs is the same**), whether it's one node or two nodes(H20 * 8). But I can run ```test_intranode.py``` normally. Could you please help me with this? thanks for your excellent job. ### Environment Docker container from https://github.com/sgl-project/sglang/blob/main/docker/Dockerfile.deepep #### nvidia-smi topo -mp ``` GPU0 GPU1 GPU2 GPU3 GPU4 GPU5 GPU6 GPU7 CPU Affinity NUMA Affinity GPU NUMA ID GPU0 X NODE NODE NODE SYS SYS SYS SYS 0-47,96-143 0 N/A GPU1 NODE X NODE NODE SYS SYS SYS SYS 0-47,96-143 0 N/A GPU2 NODE NODE X NODE SYS SYS SYS SYS 0-47,96-143 0 N/A GPU3 NODE NODE NODE X SYS SYS SYS SYS 0-47,96-143 0 N/A GPU4 SYS SYS SYS SYS X NODE NODE NODE 48-95,144-191 1 N/A GPU5 SYS SYS SYS SYS NODE X NODE NODE 48-95,144-191 1 N/A GPU6 SYS SYS SYS SYS NODE NODE X NODE 48-95,144-191 1 N/A GPU7 SYS SYS SYS SYS NODE NODE NODE X 48-95,144-191 1 N/A Legend: X = Self SYS = Connection traversing PCIe as well as the SMP interconnect between NUMA nodes (e.g., QPI/UPI) NODE = Connection traversing PCIe as well as the interconnect between PCIe Host Bridges within a NUMA node PHB = Connection traversing PCIe as well as a PCIe Host Bridge (typically the CPU) PXB = Connection traversing multiple PCIe bridges (without traversing the PCIe Host Bridge) PIX = Connection traversing at most a single PCIe bridge ``` #### ibv_devinfo ``` hca_id: erdma_0 transport: eRDMA (0) fw_ver: 0.2.0 node_guid: 0216:3eff:fe43:24be sys_image_guid: 0216:3eff:fe43:24be vendor_id: 0x1ded vendor_part_id: 4223 hw_ver: 0x0 phys_port_cnt: 1 port: 1 state: PORT_ACTIVE (4) max_mtu: 4096 (5) active_mtu: 4096 (5) sm_lid: 0 port_lid: 0 port_lmc: 0x00 link_layer: Ethernet ``` #### gdrcopy_copybw ``` GPU id:0; name: NVIDIA H20; Bus id: 0000:08:00 GPU id:1; name: NVIDIA H20; Bus id: 0000:7e:00 GPU id:2; name: NVIDIA H20; Bus id: 0000:a2:00 GPU id:3; name: NVIDIA H20; Bus id: 0000:c6:00 GPU id:4; name: NVIDIA H20; Bus id: 0001:09:00 GPU id:5; name: NVIDIA H20; Bus id: 0001:7f:00 GPU id:6; name: NVIDIA H20; Bus id: 0001:a3:00 GPU id:7; name: NVIDIA H20; Bus id: 0001:c7:00 selecting device 0 testing size: 131072 rounded size: 131072 gpu alloc fn: cuMemAlloc device ptr: 7f15a7400000 map_d_ptr: 0x7f17c8238000 info.va: 7f15a7400000 info.mapped_size: 131072 info.page_size: 65536 info.mapped: 1 info.wc_mapping: 1 page offset: 0 user-space pointer:0x7f17c8238000 writing test, size=131072 offset=0 num_iters=10000 write BW: 17649.5MB/s reading test, size=131072 offset=0 num_iters=100 read BW: 712.677MB/s unmapping buffer unpinning buffer closing gdrdrv ``` #### nvshmem-info -a ``` NVSHMEM v3.2.5 Build Information: CUDA API 12040 CUDA Driver 12080 Build Timestamp Jun 9 2025 09:20:22 Build Variables NVSHMEM_DEBUG=OFF NVSHMEM_DEVEL=OFF NVSHMEM_DEFAULT_PMI2=OFF NVSHMEM_DEFAULT_PMIX=OFF NVSHMEM_DEFAULT_UCX=OFF NVSHMEM_ENABLE_ALL_DEVICE_INLINING=OFF NVSHMEM_GPU_COLL_USE_LDST=OFF NVSHMEM_IBGDA_SUPPORT=ON NVSHMEM_IBGDA_SUPPORT_GPUMEM_ONLY=OFF NVSHMEM_IBDEVX_SUPPORT=OFF NVSHMEM_IBRC_SUPPORT=ON NVSHMEM_MPI_SUPPORT=OFF NVSHMEM_NVTX=ON NVSHMEM_PMIX_SUPPORT=OFF NVSHMEM_SHMEM_SUPPORT=OFF NVSHMEM_TEST_STATIC_LIB=OFF NVSHMEM_TIMEOUT_DEVICE_POLLING=OFF NVSHMEM_TRACE=OFF NVSHMEM_UCX_SUPPORT=OFF NVSHMEM_USE_DLMALLOC=OFF NVSHMEM_USE_NCCL=OFF NVSHMEM_USE_GDRCOPY=ON NVSHMEM_VERBOSE=OFF CUDA_HOME=/usr/local/cuda GDRCOPY_HOME=/sgl-workspace/gdrcopy-2.5 LIBFABRIC_HOME=/usr/local/libfabric MPI_HOME=/usr/local/ompi NCCL_HOME=/usr/local/nccl NVSHMEM_PREFIX=/usr/local/nvshmem PMIX_HOME=/usr SHMEM_HOME=/usr/local/ompi UCX_HOME=/usr/local/ucx Standard options: NVSHMEM_VERSION false (type: bool, default: false) Print library version at startup NVSHMEM_INFO false (type: bool, default: false) Print environment variable options at startup NVSHMEM_DISABLE_NVLS false (type: bool, default: false) Disable NVLS SHARP resources for collectives, even if available for platform NVSHMEM_SYMMETRIC_SIZE 1073741824 (type: size, default: 1073741824) Specifies the size (in bytes) of the symmetric heap memory per PE. The size is implementation-defined and must be at least as large as the integer ceiling of the product of the numeric prefix and the scaling factor. The character suffixes for the scaling factor are as follows: * k or K multiplies by 2^10 (kibibytes) * m or M multiplies by 2^20 (mebibytes) * g or G multiplies by 2^30 (gibibytes) * t or T multiplies by 2^40 (tebibytes) For example, string '20m' is equivalent to the integer value 20971520, or 20 mebibytes. Similarly the string '3.1M' is equivalent to the integer value 3250586. Only one multiplier is recognized and any characters following the multiplier are ignored, so '20kk' will not produce the same result as '20m'. Usage of string '.5m' will yield the same result as the string '0.5m'. An invalid value for NVSHMEM_SYMMETRIC_SIZE is an error, which the NVSHMEM library shall report by either returning a nonzero value from nvshmem_init_thread or causing program termination. NVSHMEM_DEBUG "" (type: string, default: "") Set to enable debugging messages. Optional values: VERSION, WARN, INFO, ABORT, TRACE Bootstrap options: NVSHMEM_BOOTSTRAP "PMI" (type: string, default: "PMI") Name of the default bootstrap that should be used to initialize NVSHMEM. Allowed values: PMI, MPI, SHMEM, plugin, UID NVSHMEM_BOOTSTRAP_PMI "PMI" (type: string, default: "PMI") Name of the PMI bootstrap that should be used to initialize NVSHMEM. Allowed values: PMI, PMI-2, PMIX NVSHMEM_BOOTSTRAP_PLUGIN "" (type: string, default: "") Absolute path to or name of the bootstrap plugin file to load when NVSHMEM_BOOTSTRAP=plugin is specified NVSHMEM_BOOTSTRAP_MPI_PLUGIN "nvshmem_bootstrap_mpi.so.3" (type: string, default: "nvshmem_bootstrap_mpi.so.3") Absolute path to or name of the MPI bootstrap plugin file. NVSHMEM will search for the plugin based on linux linker priorities. See man dlopen NVSHMEM_BOOTSTRAP_SHMEM_PLUGIN "nvshmem_bootstrap_shmem.so.3" (type: string, default: "nvshmem_bootstrap_shmem.so.3") Absolute path to or name of the SHMEM bootstrap plugin file. NVSHMEM will search for the plugin based on linux linker priorities. See man dlopen NVSHMEM_BOOTSTRAP_PMI_PLUGIN "nvshmem_bootstrap_pmi.so.3" (type: string, default: "nvshmem_bootstrap_pmi.so.3") Absolute path to or name of the PMI bootstrap plugin file. NVSHMEM will search for the plugin based on linux linker priorities. See man dlopen NVSHMEM_BOOTSTRAP_PMI2_PLUGIN "nvshmem_bootstrap_pmi2.so.3" (type: string, default: "nvshmem_bootstrap_pmi2.so.3") Absolute path to or name of the PMI-2 bootstrap plugin file. NVSHMEM will search for the plugin based on linux linker priorities. See man dlopen NVSHMEM_BOOTSTRAP_PMIX_PLUGIN "nvshmem_bootstrap_pmix.so.3" (type: string, default: "nvshmem_bootstrap_pmix.so.3") Absolute path to or name of the PMIx bootstrap plugin file. NVSHMEM will search for the plugin based on linux linker priorities. See man dlopen NVSHMEM_BOOTSTRAP_UID_PLUGIN "nvshmem_bootstrap_uid.so.3" (type: string, default: "nvshmem_bootstrap_uid.so.3") Absolute path to or name of the UID bootstrap plugin file. NVSHMEM will search for the plugin based on linux linker priorities. See man dlopen Additional options: NVSHMEM_CUDA_PATH "" (type: string, default: "") Path to directory containing libcuda.so (for use when not in default location) NVSHMEM_DEBUG_ATTACH_DELAY 0 (type: int, default: 0) Delay (in seconds) during the first call to NVSHMEM_INIT to allow for attaching a debuggger (Default 0) NVSHMEM_DEBUG_FILE "" (type: string, default: "") Debugging output filename, may contain %h for hostname and %p for pid NVSHMEM_MAX_TEAMS 32 (type: long, default: 32) Maximum number of simultaneous teams allowed NVSHMEM_MAX_MEMORY_PER_GPU 137438953472 (type: size, default: 137438953472) Maximum memory per GPU NVSHMEM_DISABLE_CUDA_VMM false (type: bool, default: false) Disable use of CUDA VMM for P2P memory mapping. By default, CUDA VMM is enabled on x86 and disabled on P9. CUDA VMM feature in NVSHMEM requires CUDA RT version and CUDA Driver version to be greater than or equal to 11.3. NVSHMEM_DISABLE_P2P false (type: bool, default: false) Disable P2P connectivity of GPUs even when available NVSHMEM_IGNORE_CUDA_MPS_ACTIVE_THREAD_PERCENTAGE false (type: bool, default: false) When doing Multi-Process Per GPU (MPG) run, full API support is available only if sum of CUDA_MPS_ACTIVE_THREAD_PERCENTAGE of processes running on a GPU is <= 100%. Through this variable, user can request NVSHMEM runtime to ignore the active thread percentage and allow full MPG support. Users enable it at their own risk as NVSHMEM might deadlock. NVSHMEM_CUMEM_GRANULARITY 536870912 (type: size, default: 536870912) Granularity for cuMemAlloc/cuMemCreate NVSHMEM_PROXY_REQUEST_BATCH_MAX 32 (type: int, default: 32) Maxmum number of requests that the proxy thread processes in a single iteration of the progress loop. Collectives options: NVSHMEM_DISABLE_NCCL false (type: bool, default: false) Disable use of NCCL for collective operations NVSHMEM_BARRIER_DISSEM_KVAL 2 (type: int, default: 2) Radix of the dissemination algorithm used for barriers NVSHMEM_BARRIER_TG_DISSEM_KVAL 2 (type: int, default: 2) Radix of the dissemination algorithm used for thread group barriers NVSHMEM_FCOLLECT_LL_THRESHOLD 2048 (type: size, default: 2048) Message size threshold up to which fcollect LL algo will be used NVSHMEM_REDUCE_SCRATCH_SIZE 524288 (type: size, default: 524288) Amount of symmetric heap memory (minimum 16B, multiple of 8B) reserved by runtime for every team to implement reduce and reducescatter collectives NVSHMEM_BCAST_ALGO 0 (type: int, default: 0) Broadcast algorithm to be used. * 0 - use default algorithm selection strategy NVSHMEM_REDMAXLOC_ALGO 1 (type: int, default: 1) Reduction algorithm to be used for MAXLOC operation. * 1 - default, flag alltoall algorithm * 2 - flat reduce + flat bcast * 3 - topo-aware two-level reduce + topo-aware bcast Transport options: NVSHMEM_REMOTE_TRANSPORT "ibrc" (type: string, default: "ibrc") Selected transport for remote operations: ibrc, ucx, libfabric, ibdevx, none NVSHMEM_ENABLE_NIC_PE_MAPPING false (type: bool, default: false) When not set or set to 0, a PE is assigned the NIC on the node that is closest to it by distance. When set to 1, NVSHMEM either assigns NICs to PEs on a round-robin basis or uses NVSHMEM_HCA_PE_MAPPING or NVSHMEM_HCA_LIST when they are specified. NVSHMEM_DISABLE_LOCAL_ONLY_PROXY false (type: bool, default: false) When running on an NVLink-only configuaration (No-IB, No-UCX), completely disable the proxy thread. This will disable device side global exit and device side wait timeout polling (enabled by NVSHMEM_TIMEOUT_DEVICE_POLLING build-time variable) because these are processed by the proxy thread. NVSHMEM_IB_ENABLE_IBGDA false (type: bool, default: false) Set to enable GPU-initiated communication transport. Hidden options: NVSHMEM_INFO_HIDDEN true (type: bool, default: false) Print hidden environment variable options at startup NVSHMEM_DISABLE_NVLS_SHARING true (type: bool, default: true) Disable NVLS SHARP resource sharing for user-defined teams NVSHMEM_HEAP_KIND "DEVICE" (type: string, default: "DEVICE") Specify the memory kind used by the NVSHMEM symmetric heap. Allowed values: VIDMEM, SYSMEM NVSHMEM_ENABLE_RAIL_OPT false (type: bool, default: false) Enable Rail Optimization when heap is in SYSMEM NVSHMEM_BOOTSTRAP_TWO_STAGE false (type: bool, default: false) Ignore CUDA device setting during initialization,forcing two-stage initialization NVSHMEM_DEBUG_SUBSYS "" (type: string, default: "") Comma separated list of debugging message sources. Prefix with '^' to exclude. Values: INIT, COLL, P2P, PROXY, TRANSPORT, MEM, BOOTSTRAP, TOPO, UTIL, ALL NVSHMEM_ENABLE_ERROR_CHECKS false (type: bool, default: false) Enable error checks NVSHMEM_DISABLE_MNNVL false (type: bool, default: false) Disable MNNVL connectivity for GPUs even when available NVSHMEM_CUMEM_HANDLE_TYPE "FILE_DESCRIPTOR" (type: string, default: "FILE_DESCRIPTOR") Handle type for cuMemCreate. Supported are - FABRIC or FILE_DESCRIPTOR NVSHMEM_BYPASS_ACCESSIBILITY_CHECK false (type: bool, default: false) Bypass peer GPU accessbility checks NVSHMEM_FCOLLECT_NTHREADS 512 (type: int, default: 512) Sets number of threads per block for fcollect collective. By default, if no env is set, default value is min(max_occupancy per CTA, msg size per PE). If env is specified, value overrides the default irrespective of max occupancy per CTA NVSHMEM_REDUCESCATTER_NTHREADS 512 (type: int, default: 512) Sets number of threads per block for reducescatter collective. By default, if no env is set, default value is min(max_occupancy per CTA, msg size per PE). If env is specified, value overrides the default irrespective of max occupancy per CTA NVSHMEM_MAX_CTAS 1 (type: int, default: 1) Sets number of blocks per grid for host onstream collective. By default, if no env is set, default value to 1 CTA If env is specified, value overrides the default value NVSHMEM_REDUCE_RECEXCH_KVAL 2 (type: int, default: 2) Radix of the recursive exchange reduction algorithm NVSHMEM_FCOLLECT_LL128_THRESHOLD 0 (type: size, default: 0) Message size threshold up to which the fcollect LL128 algo will be used. LL128 will be used only when FCOLLECT_LL_THRESHOLD < size NVSHMEM_FCOLLECT_NVLS_THRESHOLD 16777216 (type: size, default: 16777216) Message size threshold up to which fcollect NVLS algo will be used NVSHMEM_REDUCESCATTER_NVLS_THRESHOLD 16777216 (type: size, default: 16777216) Message size threshold up to which reducescatter NVLS algo will be used NVSHMEM_BCAST_TREE_KVAL 2 (type: int, default: 2) Radix of the broadcast tree algorithm NVSHMEM_FCOLLECT_ALGO 0 (type: int, default: 0) Fcollect algorithm to be used. * 0 - use default algorithm selection strategy NVSHMEM_REDUCE_ALGO 0 (type: int, default: 0) Allreduce algorithm to be used. * 0/1 - use default algorithm selection strategy NVSHMEM_REDUCE_NVLS_THRESHOLD 2048 (type: int, default: 2048) Message size threshold up to which allreduce one-shot algo will be used NVSHMEM_REDUCESCATTER_ALGO 0 (type: int, default: 0) Reduce Scatter algorithm to be used. * 0 - use default algorithm selection strategy NVSHMEM_ASSERT_ATOMICS_SYNC false (type: bool, default: false) Bypass flush on wait_until at target NVSHMEM_BYPASS_FLUSH false (type: bool, default: false) Bypass flush in proxy when enforcing consistency NVTX options: NVSHMEM_NVTX "off" (type: string, default: "off") Set to enable NVTX instrumentation. Accepts a comma separated list of instrumentation groups. By default the NVTX instrumentation is disabled. init : library setup alloc : memory management launch : kernel launch routines coll : collective communications wait : blocking point-to-point synchronization wait_on_stream : point-to-point synchronization (on stream) test : non-blocking point-to-point synchronization memorder : memory ordering (quiet, fence) quiet_on_stream : nvshmemx_quiet_on_stream atomic_fetch : fetching atomic memory operations atomic_set : non-fetchong atomic memory operations rma_blocking : blocking remote memory access operations rma_nonblocking : non-blocking remote memory access operations proxy : activity of the proxy thread common : init,alloc,launch,coll,memorder,wait,atomic_fetch,rma_blocking,proxy all : all groups off : disable all NVTX instrumentation ``` ### The error is as follows #### 1 node: ``` root@zz:/sgl-workspace/DeepEP# python3 tests/test_low_latency.py Allocating buffer size: 2115.111296 MB ... WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping... /sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping... /sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping... /sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping... /sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping... /sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping... /sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping... /sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA WARN: device erdma_0 is not enumerated as an mlx5 device. Skipping... /sgl-workspace/nvshmem/src/host/transport/transport.cpp:nvshmemi_transport_init:275: init failed for transport: IBGDA /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value srq creation failed srq creation failed srq creation failed srq creation failed /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:1395: non-zero status: 7 /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:1395: non-zero status: 7 /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:1395: non-zero status: 7 /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:1395: non-zero status: 7 ep_create failed ep_create failed ep_create failed ep_create failed [iZ2ze1lasox0u2bv6k3iicZ:2136 :0:2136] Caught signal 11 (Segmentation fault: address not mapped to object at address 0x38) [iZ2ze1lasox0u2bv6k3iicZ:2137 :0:2137] Caught signal 11 (Segmentation fault: address not mapped to object at address 0x38) [iZ2ze1lasox0u2bv6k3iicZ:2142 :0:2142] Caught signal 11 (Segmentation fault: address not mapped to object at address 0x38) [iZ2ze1lasox0u2bv6k3iicZ:2135 :0:2135] Caught signal 11 (Segmentation fault: address not mapped to object at address 0x38) /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp:318: NULL value srq creation failed srq creation failed srq creation failed srq creation failed ``` When I add ```NVSHMEM_DEBUG=1```, it occurs some info (the log is too long so I show the abnormal ones): ``` /sgl-workspace/nvshmem/src/modules/transport/common/transport_gdr_common.cpp 73 GDR driver version: (2, 5) /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp 1632 Begin - Enumerating IB devices in the system ([<dev_id, device_name, num_ports>]) - /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp 1652 Enumerated IB devices in the system - device id=0 (of 1), name=erdma_0, num_ports=1 /sgl-workspace/nvshmem/src/modules/transport/common/transport_ib_common.h 103 NVSHMEM_IB_ADDR_FAMILY set by environment to AF_INET /sgl-workspace/nvshmem/src/modules/transport/common/transport_ib_common.h 126 NVSHMEM_IB_ADDR_RANGE set by environment to ::/0 /sgl-workspace/nvshmem/src/modules/transport/common/transport_ib_common.h 138 NET/IB: Ip address '::' is invalid for family AF_INET, ignoring address /sgl-workspace/nvshmem/src/modules/transport/ibrc/ibrc.cpp 1739 End - Enumerating IB devices in the system ``` ``` /sgl-workspace/nvshmem/src/modules/transport/ibgda/ibgda.cpp 3885 no active IB device that supports GPU-initiated communication is found, exiting... ``` #### 2 nodes: There is none output and hang until over time. ``` root@zz:/sgl-workspace/DeepEP#MASTER_ADDR= MASTER_ADDR WORLD_SIZE=2 RANK=0 python3 tests/test_internode.py ``` When I use py-spy to output the stacks of all threads, it shows the thread hangs in init buffer of deepep: ``` root@zz:py-spy --dump -p 7422 Python v3.10.12 (/usr/bin/python3.10) Thread 7422 (active): "MainThread" __torch_function__ (torch/utils/_device.py:104) all_gather_object (torch/distributed/distributed_c10d.py:3036) wrapper (torch/distributed/c10d_logger.py:81) __init__ (deep_ep/buffer.py:66) test_loop (test_internode.py:229) _wrap (torch/multiprocessing/spawn.py:90) run (multiprocessing/process.py:108) _bootstrap (multiprocessing/process.py:314) _main (multiprocessing/spawn.py:129) spawn_main (multiprocessing/spawn.py:116) <module> (<string>:1) ``` These problems originated from my use of sglang PD disggeraton to run deepseek. I found that enabling deepep would result in the inability to use multiple P nodes or D nodes, and the inability to enable the low_latency mode of the D node. For now, Deepep’s problem cannot be solved. Is there anything reason?

bsakaguchi closed this issue 2026-10-26 09:55:30 +00:00

bsakaguchi commented 2026-10-26 09:55:31 +00:00

Author

Owner

Copy link

@sphish commented on GitHub (Jun 10, 2025):

As far as I know, eRDMA does not currently support IBGDA. You can consult your cloud service provider for more information.

<!-- gh-comment-id:2958481980 --> @sphish commented on GitHub (Jun 10, 2025): As far as I know, eRDMA does not currently support IBGDA. You can consult your cloud service provider for more information.

bsakaguchi commented 2026-10-26 09:55:31 +00:00

Author

Owner

Copy link

@Tyndalllll commented on GitHub (Jun 10, 2025):

As far as I know, eRDMA does not currently support IBGDA. You can consult your cloud service provider for more information.

Thanks, I learned that the current official version of erdma does not support IBGDA, but it will be supported in the near future.

So I think this problem can be closed and I will test deepep after the new edrma version is released.

<!-- gh-comment-id:2958841503 --> @Tyndalllll commented on GitHub (Jun 10, 2025): > As far as I know, eRDMA does not currently support IBGDA. You can consult your cloud service provider for more information. Thanks, I learned that the current official version of erdma does not support IBGDA, but it will be supported in the near future. So I think this problem can be closed and I will test deepep after the new edrma version is released.

Sign in to join this conversation.

No Branch/Tag specified

Branches Tags

main

hybrid-ep

chore/remove-clang-format-yapf

v2.5

nvDev

try-fix-cu128

mori-ep

antgroup-opt

compat-old-torch

add-linter

eager-rdma

v1.2.1

Labels

Clear labels   

pull-request

Mirrored from GitHub Pull Request

No labels

pull-request

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

deepseek-ai/DeepEP#165

No description provided.