Removing uplinks from an Uplink Profile causes incorrect TEP order and some VMs lose connectivity
Article ID: 345785
Updated On:
Products
Issue/Introduction
This is meant to provide guidance on the workflow required to reduce the number of uplinks in an Uplink Profile for NSX versions prior to 3.2.2.
Symptoms:
- After reducing the number of uplinks in an Uplink Profile, some VMs lose network connectivity.
- Adding the uplinks back to the Uplink Profile restores network connectivity.
- Logging in /var/run/log/nsx-syslog.log shows an update with incorrect TEP order.
Example logging with correct order for the 4 TEPs, vmk10, vmk11, vmk12, vmk13:
2022-11-03T23:19:51Z nsx-opsagent[6348138]: NSX 6348138 - [nsx@6876 comp="nsx-esx" subcomp="opsagent" s2comp="nsxa" tid="6348171" level="INFO"] Upgrading - restoring backup HostConfigMsg [host_switches { switch_id: "<UUID>" host_switch_profiles { uplink_profile { teaming_policy: LOADBALANCE_PORT_ID active { type: PHYSICAL_NIC name: "uplink-1" } active { type: PHYSICAL_NIC name: "uplink-2" } active { type: PHYSICAL_NIC name: "uplink-3" } active { type: PHYSICAL_NIC name: "uplink-4" } } } pnics { device_name: "uplink1" uplink_name: "uplink-1" } pnics { device_name: "uplink2" uplink_name: "uplink-2" } pnics { device_name: "uplink3" uplink_name: "uplink-3" } pnics { device_name: "uplink4" uplink_name: "uplink-4" } vteps { ip: "<vmk10 IP>" device: "" operation: ADD } vteps {
2022-11-03T23:19:51Z nsx-opsagent[6348138]: ip: "<vmk11 IP>" device: "" operation: ADD } vteps { ip: "<vmk12 IP>" device: "" operation: ADD } vteps { ip: "<vmk13 IP>" device: "" operation: ADD }
Example update where order became vmk10, vmk12, vmk11, vmk13:
2022-11-03T23:23:42Z nsx-opsagent[6348138]: NSX 6348138 - [nsx@6876 comp="nsx-esx" subcomp="opsagent" s2comp="nsxa" tid="6348167" level="INFO"] [ProcessReceiveMessages] HostConfigMsg [host_switches { switch_id: "<UUID>" host_switch_profiles { uplink_profile { teaming_policy: LOADBALANCE_PORT_ID active { type: PHYSICAL_NIC name: "uplink-1" } active { type: PHYSICAL_NIC name: "uplink-2" } active { type: PHYSICAL_NIC name: "uplink-3" } active { type: PHYSICAL_NIC name: "uplink-4" } } } pnics { device_name: "uplink1" uplink_name: "uplink-1" } pnics { device_name: "uplink2" uplink_name: "uplink-2" } pnics { device_name: "uplink3" uplink_name: "uplink-3" } pnics { device_name: "uplink4" uplink_name: "uplink-4" } vteps { ip: "<vmk10 IP>" device: "vmk10" operation: UPDATE } vteps {
2022-11-03T23:23:42Z nsx-opsagent[6348138]: ip: "<vmk12 IP>" device: "vmk12" operation: UPDATE } vteps { ip: "<vmk11 IP>" device: "vmk11" operation: UPDATE } vteps { ip: "<vmk13 IP>" device: "vmk13" operation: UPDATE }
Cause
The Management plane fetches existing TEPs from a hash map. Previously, the TEP order was not properly maintained in that hash map when making changes to Uplink Profile.
Resolution
NSX-INST-UPLINK
This issue is resolved in NSX version 3.2.2.
Workaround:
To reduce uplink count in Uplink Profile on versions before 3.2.2:
1. Place host into Maintenance Mode
2. Configure a new Uplink Profile with the reduced uplink count
3. Uninstall NSX from the host
4. Reconfigure NSX on the host with the updated Uplink Profile
Additional Information
If TEP order is atypical (ex: vmk10, 12, 11, 13) when reducing the uplink count in an Uplink Profile, VMs now using TEPs that were expected to have been deleted (vmk12 in above example when going from 4 uplinks to 2 uplinks) lose network connectivity.
Feedback
