[SX-Aurora TSUBASA] InfiniBand for SX-Aurora TSUBASA FAQ
Question
About Installation
-
Installing ve-infiniband group fails with following messages.
Error: Package: ve-memory-mapping-X.X.X-X.el7.x86_64 (TSUBASA-repo)
Requires: ve_peermem
Or
Error:
Problem: package ve-memory-mapping-X.X.X-X.el8.x86_64 requires ve_peermem, but none of the providers can be installed
- package kmod-ve_peermem-mofedX.X-X.X.X-X.el8.x86_64 is filtered out by exclude filtering
-
Installing ve-infiniband group fails with following messages.
Error: Package: kmod-ve_peermem-mofedX.X-X.X.X-X.el7.x86_64 (TSUBASA-repo)
Requires: ksym(ib_register_peer_memory_client) = 0xXXXXXXXX
Or
Error:
Problem 1: conflicting requests
- nothing provides ksym(ib_register_peer_memory_client) = 0xXXXXXXXX needed by kmod-ve_peermem-mofedX.X-X.X.X-X.el8.x86_64
-
Installing ve-infiniband group fails with following messages.
Error: Package: kmod-ve_peermem-mofedX.X-X.X.X-X.el7.x86_64 (TSUBASA-repo)
Requires: kernel(XXXXXXXXXXXX) = 0xXXXXXXXX
Or
Error:
Problem 1: conflicting requests
- nothing provides kernel(XXXXXXXXXXXX) = 0xXXXXXXXX needed by kmod-ve_peermem-mofedX.X-X.X.X-X.el8.x86_64
-
While installing following message is displayed.
depmod: WARNING: /lib/modules/kernel-version/extra/ve_peermem/ve_peermem.ko needs unknown symbol ib_register_peer_memory_client
About NEC MPI
-
Executing NEC MPI fails with following messages.
[0] MPID_OFED_Open_hca: open device failed ib_dev 0xXXXXXXXXXXXX name mlx5_X
[0] Error in Infiniband/OFED initialization. Execution aborts
Answer
About Installation
-
Please check exclude option in /etc/yum.conf and /etc/dnf/dnf.conf.
If kmod-* is specified, please remove it.
-
The followings are the possibilities:
- MLNX_OFED was installed before updating the kernel, or
- MLNX_OFED was installed with --add-kernel-support option, but --kmp option was not specified.
Please re-install MLNX_OFED.
Please refer to "2.3 Installation of MLNX_OFED (Optional)" in "SX-Aurora TSUBASA Installation Guide".
-
The kernel in use may not be supported by SX-Aurora TSUBASA.
Please refer to the following and use the kernel which is supported by SX-Aurora TSUBASA.
[SX-Aurora TSUBASA]Verified Linux kernel
-
This message may be displayed when multiple kernels are installed.
Please ignore the message if the displayed kernel version is not matched to the kernel in use.
About NEC MPI
-
There is a possibility that MLNX_OFED and ve-infiniband are not re-installed after updating the kernel.
Re-installing MLNX_OFED and ve-infiniband is mandatory after updating the kernel.
Please retry "Chapter5 Update" in "SX-Aurora TSUBASA Installation Guide".
Product Name
SX-Aurora TSUBASA Software
Note
Update History
2023/09/28 Update links
Update Mellanox OFED to MLNX_OFED
2021/09/30 New Release
-
Content ID:
4150101112
-
Release date:
2021/09/30
-
Last updated:2023/09/28