Network/routing issues caused by using LACP bond with k8s in SCALE
Description
Problem/Justification
None
Impact
None
SmartDraw Connector
Katalon Manual Tests (BETA)
Activity
Show:
Bug Clerk May 19, 2021 at 6:14 PM
Complete
Pinned fields
Click on the next to a field label to start pinning.
Details
Details
Assignee
Waqar AhmedReporter
Nick PrincipeImpact
High
Components
Fix versions
Affects versions
Priority
MediumMore fields
Time tracking
More fields
Time trackingKatalon Platform
Linked Test Cases, Katalon Defect Results, Katalon Studio Test Results
Katalon Platform
Linked Test Cases, Katalon Defect Results, Katalon Studio Test Results
Created May 3, 2021 at 8:16 PM
Updated July 6, 2022 at 9:00 PM
Resolved May 20, 2021 at 10:54 AM
When using an LACP bond with an app in k8s in SCALE 21.04-ALPHA1, and I believe even newer nightlies, I encounter routing issues on my network where one of the SCALE controllers with bond0 and k8s tries to become a router and reply to ICMP echo requests for the other SCALE node that also has bond0 and k8s.
I believe that k8s + LACP bond0 is required to reproduce this. I am not sure if two controllers both with k8s + LACP are required, but believe that may be the case.
I narrowed this down by destroying the bond0 on both SCALE controllers and replacing with a single non-LACP interface. After doing this, the k8s app cluster came up and all four nodes in the cluster could then reliably communicate with each other.
Right now it seems that it could be some misconfiguration that is causing MAC address/ARP/STP issues.
In my case when I was trying to use an LACP bond, it was a 4 interface LACP group (2x dual port 100GbE NIC) in each controller. No other interfaces on the system were configured at the time (except IPMI, invisible to the OS.)