Bug 1637389

Description of problem:

Intermittent problem with network connection loss.
The problems do *not* occur and have *never*
occurred when accessing the mongo or postgres pods from the lb
pod.

The connectivity problems occur only when customer is  accessing mongo or
postgres pods from the app pod. Due to those connectivity problems
the app pod fails after a short time, so customer will have to re-deploy
it first and then do the tests before it fails and gets restarted.

Version-Release number of selected component (if applicable):
3.9.41

How reproducible:
Customer send us a video with this problem :
https://www.dropbox.com/s/ivpowk4iewetc1n/RedHatSupportCase02168013.mp4

Steps to Reproduce:
1. While the app-a application is down, connecting repeatedly
from the lb-0 pod to postgres-1, which works as expected
2. Deploying the app-a application 
3. Waiting for app-a to start up, then connecting repeatedly
from app-a to postgres-1. Here about half the connect
attempts trigger 'Network is unreachable' errors in a
seemingly random pattern 
4. Connecting again from lb-0 while app-a is still running,
no problems observed here 

Actual results:
App 

Expected results:


Additional info: