ECMP load-balancing using floating ip is not working correctly
Bug #1687879 reported by
Vedamurthy Joshi
This bug affects 2 people
Affects | Status | Importance | Assigned to | Milestone | ||
---|---|---|---|---|---|---|
Juniper Openstack | Status tracked in Trunk | |||||
R3.2 |
Fix Committed
|
High
|
Hari Prasad Killi | |||
R4.0 |
Fix Committed
|
High
|
Hari Prasad Killi | |||
Trunk |
Fix Committed
|
High
|
Hari Prasad Killi |
Bug Description
R4.0 3060
kubernetes setup
A service has a floating ip 10.111.84.52 with 5 backend pods
Two are in one compute node(CN1) and 3 are on the other compute node(CN2)
Client pod is on CN1 doing a http request to the floating IP
It is seen that load balancing is happening only on 3 out of the 5 pods.
On the source compute, load-balancing is fine, but across the tunnel, traffic always goes to the first ecmp index
tags: | added: dt |
To post a comment you must log in.
This issue was still seen during 4.0 build 10 test run
It was seen that this time, traffic need not fail by going to the first ecmp index only though, but could go to other indices also.