search

stanford-rc / ibswinfo

Command-line tool to retrieve information and monitor Mellanox un-managed Infiniband switches
GNU General Public License v3.0
49 stars 10 forks source link

SwitchIB support and issues #5

Closed keenandr closed 4 years ago

keenandr commented 4 years ago

Hi,

The unmanaged EDR Switch-IB based switches (SB7790) appear to mostly work. The PSID is reported incorrectly; it appears to be transposed?

# flint -d /dev/mst/SW_MT52000_SwitchIB_Mellanox_Technologies_lid-0x0005 q | grep PSID
PSID:                  HP_1880110032

versus

# ./ibswinfo.sh -d lid-5 | grep PSID
PSID               | 1_PH108830012

It also only reports 35 instead of 36 ports.

kcgthb commented 4 years ago

Thanks for reporting this!

I unfortunately don't have any SB7790 model to test directly, but the first thing that comes to mind is that HP_1880110032 is not a Mellanox PSID, but rather an HPE one. It's possible that the HPE firmware for those switches behaves differently, in ways that could result in the register addresses being shifted.

Could you send me the output of the following commands?

# mlxreg -d lid-5 --reg_name MGIR --get
# mlxreg -d lid-5 --reg_name MGPIR --get
# smpquery NI 5
kcgthb commented 4 years ago

Thank you for the details you provided by email! Both issues should be fixed in 0.3