PowerFlex: ESXi PNA deployments with OS10 switches may fail depending on the switch cabling
Oversigt: PowerFlex Manager installs operating systems during new service deployment or node expansion with Partial Network Automation using iDRAC virtual media. Sometimes when LACP networking is used with Dell OS 10 switches OS installation failures can be seen depending on the switch cabling used. ...
Symptomer
In /opt/Dell/ASM/deployments/<service_ID>/ log file, you may see something similar to this:
Årsag
For ESXi, LACP has to be configured after the node is added to vCenter. Prior to that the ESXi node will use only a single vmnic. This configuration is generally not compatible with LACP networking, but the "LACP fallback" option can be used to allow it to function. This feature allows the port channel to gracefully degrade to individual Ethernet ports when LACP traffic is not seen as is the case during PXE OS Installation. PowerFlex Manager uses the "LACP fallback" option to allow ESXi OS installation to proceed when LACP networking is requested.
However on Dell OS 10 switches, when "LACP fallback" is active only one of the two switch interfaces that are part of the LACP port channel is active. This can cause ESXi OS installation with PowerFlex Manager to fail if the first NIC port used for OS installation is not connected to the switch port that is active.
This document includes more details about which switch port will be elected as active: LACP fallback in VLT | Dell SmartFabric OS10 - Virtual Link Trunking - Reference Architecture Guide | Dell Technologies Info Hub
Løsning
If Partial Network Automation OS installation fails with Dell OS 10 switches, find the switch interface connected to the first NIC port with the management network on it. Then log in to the switch, run "show port channel summary" and search for the port channel containing this switch interface. If it reports "D" for down then you may be affected by this issue. To confirm, find the switch interface connected to the second NIC port with the management network on it. Log in to the other switch and run "show port channel summary." If it reports "UF" for Up, Fallback enabled then you are affected by this issue. Note that the same switch will be the active switch for all LACP fallback-enabled port channels connected to this pair of switches.
These examples show that the switch "X-TOR4B" is the active switch for LACP fallback:
X-TOR4B# show port-channel summary
Flags: D - Down I - member up but inactive P - member up and active
U - Up (port-channel) F - Fallback Activated
--------------------------------------------------------------------------------
Group Port-Channel Type Protocol Member Ports
--------------------------------------------------------------------------------
1 port-channel1 (UF) Eth DYNAMIC 1/1/5(P)
2 port-channel2 (UF) Eth DYNAMIC 1/1/6(P)
X-TOR4A# show port-channel summary
Flags: D - Down I - member up but inactive P - member up and active
U - Up (port-channel) F - Fallback Activated
--------------------------------------------------------------------------------
Group Port-Channel Type Protocol Member Ports
--------------------------------------------------------------------------------
1 port-channel1 (D) Eth DYNAMIC 1/1/5(I)
2 port-channel2 (D) Eth DYNAMIC 1/1/6(I)
The resolution is to swap the cables for the two ports mentioned above so that the first NIC port is connected to the switch with the active interface.