User Manual

Rev 4.60
Mellanox Technologies
124
10 Troubleshooting
10.1 InfiniBand Troubleshooting
Issue 1. The InfiniBand interfaces are not up after the first reboot after the installation process is
completed.
Suggestion: To troubleshoot this issue, follow the steps below:
1. Check that the InfiniBand driver is running on all nodes by using “vstat”. The
vstat utility located at <installation_directory>\tools, displays the status and capabil-
ities of the network adaptor card(s).
2. On the command line, enter “vstat” (use -h for options) to retrieve information
about one or more adapter ports. The field port_state will be equal to:
• PORT_DOWN - when there is no InfiniBand cable ("no link");
• PORT_INITIALIZED - when the port is connected to some other port ("physical link");
• PORT_ACTIVE - when the port is connected and OpenSM is running ("logical link")
• PORT_ARMED - when the port is connected to some other port ("physical link");
3.
Run “sminfo” and verify that OpenSM is running.
In case OpenSM is not running, please see OpenSM operation instructions in
Section 7, “OpenSM - Subnet Manager”, on page 60 above.
4. Verify the status of ports by using vstat: All connected ports should report
"PORT_ACTIVE" state.
10.2 Ethernet Troubleshooting
Issue 1. The installation of Win OFED VPI for Windows fails with the following error mes-
sage:
This installation package is not supported by this processor type. Contact your product vendor."
Suggestion: This message is printed if you have downloaded and attempted to install an incor-
rect driver version-- for example, if you are trying to install a 64-bit driver on a 32-bit machine
(or vice versa).
Issue 2. The performance is low.
Suggestion: This can be due to non-optimal system configuration. See the section "Perfor-
mance Tuning" to take advantage of Mellanox 40/10 GBit NIC performance.
Issue 3. The driver does not start.
Suggestion 1: This can happen due to an RSS configuration mismatch between the TCP stack
and the Mellanox adapter. To confirm this scenario, open the event log and look under "Sys-
tem" for the "mlx4ethX" source. If found, enable RSS as follows:
1. Run the following command: "netsh int tcp set global rss = enabled".
Suggestion 2: This is a less recommended suggestion, and will cause low performance. To dis-
able RSS on the adapter, run the following command: "netsh int tcp set global rss = no dynamic
balancing".
Issue 4. The Ethernet driver fails to start. In the Event log, under the mlx4_bus source, the fol-
lowing error message appears: RUN_FW command failed with error -22