Add upstream keepalive connections cache

[aledbf]

fixes #838

[k8s-reviewable]

This change is 

[coveralls]

Coverage increased (+0.04%) to 44.551% when pulling 6e22035 on aledbf:keepalive into 4235166 on kubernetes:master.

[coveralls]

Coverage increased (+0.008%) to 44.523% when pulling 6e22035 on aledbf:keepalive into 4235166 on kubernetes:master.

[xcompass]

In case someone run into the same problem as mine: I have a spring boot application, when the keepalive is enabled, the application respond to a few requests then stopped to respond. If I set keepalive to 1, the application stops responding for every other request. Had to remove keepalive option. No problem with other apps.

[necevil]

@xcompass I think I may be running into a similar issue, how did you remove the keep alive option for the upstream service?

[xcompass]

@necevil I don't have that app anymore, so I can't check. But if I remember correctly, I had to remove the keepalive option from Nginx config (disabled by default according to this PR). You can check your nginx config to see if there is an keepalive option enabled. Hope it helps.