Dell Emulex Family of Adapters - Manual (Page 1132)

Emulex Drivers Version 10.2 for VMware ESXi User Manual P010087-01A Rev. A
4. Troubleshooting
Troubleshooting the NIC Driver
1132
Troubleshooting the NIC Driver
The following section includes NIC troubleshooting information. Be sure to check the
readme.txt file located on CD1 for other troubleshooting issues.
The following table provides troubleshooting information for the NIC driver.
Table 4-4 Troubleshooting the NIC Driver
Issue Resolution
1) When there is a great deal of
network traffic in some VMs, a few
VMs appear to have lost network
connectivity.
2) Several “alloc_skb() failed”
messages appear in the log file:
/proc/vmware/log
This could be due to low configured value for
netPktHeapMaxSize. Try increasing it to a higher value. To
read the current value, run:
# esxcfg-advcfg -j netPktHeapMaxSize
(A value of 0 indicates default - 64 MB) To increase the size
to (for example, 128 MB), run:
# esxcfg-advcfg -k 128 netPktHeapMaxSize
(netPktHeapMaxSize can also be configured through VI Client
using Configuration > Advanced Settings > VMKernel.)
After configuring the size, reboot the system.
Unable to ping from one VM to another
VM. The NIC driver creates two vmnic interfaces, one for each
port. If these interfaces are configured as uplinks in two
separate vSwitches, the VMs in each of these switches are in
separate networks with no network path between them.
Thus, pinging between the VMs in the two groups fails. If you
want all these VMs in the same network, configure them as
teaming uplinks to one vSwitch option. Each of the vmnics,
vmnic1 to vmnic16, must be configured in a separate
vSwitch. In this configuration, there is no network path
between the vSwitches and pinging between these VMs does
not work.
When inserting or removing a 1Gb
SFP-RJ45 module on an adapter
without RJ45 copper cables attached,
the operating system indicates link up
or down status. When inserting or
removing copper cables attached to a
switch to the module, link up or down
events are not reported to the
operating system. There is no solution to this issue as the link status is not
reported due to a PHY limitation on the card.
Flow control setting is not stored per
port after rebooting the system. With flow control, there is no persistence across reboot. It
always starts with both RX and TX on. For persistence, run a
config command from an RC file at reboot.