ipv6 addresses not automatically wrapped

[wombat]

What did you do?
We installed Kubernetes and Prometheus on top of Rackspace and seeing sometimes ipv4, sometimes ipv6 addresses for nodes, service and pod ips.

What did you expect to see?
All endpoints either declared explicitly or via autodiscovery should be in state UP

What did you see instead? Under which circumstances?
Endpoints with a ipv6 addresses have a state DOWN and the error says Get http://xxxx:xxxx:xxxx:xxx:xxxx:xxxx:xxxx:xxx:9100/metrics: dial tcp: too many colons in address xxxx:xxxx:xxxx:xxx:xxxx:xxxx:xxxx:xxx:9100

Environment

• System information:
  Linux 4.7.0-coreos x86_64
• Prometheus version:
  prometheus, version 1.1.0 (branch: master, revision: a60378c)
  build user: root@583368332c21
  build date: 20160903-19:35:49
  go version: go1.6.3
• Alertmanager version:

none

• Prometheus configuration file:

https://raw.githubusercontent.com/prometheus/prometheus/v1.1.0/documentation/examples/prometheus-kubernetes.yml

• Alertmanager configuration file:
  none
• Logs:
  none
• Screens:
  

[fabxc]

I see where we are rendering them wrong in the UI. However, for scraping we are using URL.String(), which handles ipv6 addresses correctly.

[jimmidyson]

@fabxc Aren't we using IP.String() which doesn't handle it properly? It appears it doesn't... https://play.golang.org/p/Q8ikNMh4aF Haven't looked properly though...

[jimmidyson]

e.g. https://github.com/prometheus/prometheus/blob/master/retrieval/discovery/kubernetes/node.go#L95

[fabxc]

Ah, yes. Just walked to the Prometheus code. Quite a few places where unexpected things could happen with ipv6. Will work on a fix.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.