Open LionelJouin opened 2 years ago
Here is new logs: traffic-disturbance.zip
The tests passed 8 times and failed on attempt 9.
Logs in the network
directory have been collected after scaling to 5 and just before sending traffic that have failed to reach the 5 targets. At that time, the load-balancers can ping and send traffic to every TCP traffic to all targets (not shown in the logs, but I tried it).
Here are targets that have received traffic:
target-a-5cbdfc758-kc8dr
target-a-5cbdfc758-pqdns
target-a-5cbdfc758-x9nhq
And the ones that haven't:
target-a-5cbdfc758-8m6zm
target-a-5cbdfc758-hkcx8
Describe the bug The issue is similar to this one: https://github.com/Nordix/Meridio/issues/55 After scaling the targets, for instance from 4 to 5, some targets are not receiving traffic for x seconds while they are correctly configured in nfqlb with correct IP rules and IP routes. It's random, most of the time it works correctly, sometimes only 1 is receiving traffic, sometimes 2.
I tried with ctraffic and mconnect, the result is the same.
To Reproduce
Expected behavior All targets should receive the traffic.
Context
Running on Kind, I haven't tried any other environment, but it happened to the internal CI team.
Logs