Merge pull request atlassian#144 from skydiator/master

atlassian#145: Ensure min_nodes and max_nodes are up to date with the cloud provider.

upgrade client-go to v10.0.0 (atlassian#141)

* upgrade client-go to v10.0.0 and go version to 1.10+

escalator/60 - Added metrics histogram of how long new nodes take to …

…register in kube from cloud provider resource instantiation. (atlassian#137)

fix node not belong to ASG issue when scaling down (atlassian#138)

* error out the app when node not found inside expected ASG and move Fatal out of library into main

* change errors library to be consistent everywhere

* fix tests after updating errors library

* update Gopkg dependencies

* move node not in group error into cloudprovider and update error check logic for it inside controller

* migrate more Fatal error processing into main function

* address some comments

Add scale lock summary metric to observe lock durations & count (atla…

…ssian#134)

* escalator/60:
  - Added some Makefile wiring.
  - Added scale lock duration counter to experiment with prom vectors.
  - Plumbing the Node Group name to the State Lock; using this to tag lock duration metrics on unlock.
  - Scale lock `unlock()` noop if already unlocked, stamping node group name on scale lock from within controller.
  - Working on improving turnaround of the docker build.
  - Changed lock duration metric to summary type and storing value as seconds.

* escalator/60:
  - `node_group_scale_lock` metric now a gauge. Moved in to `scale_lock.go lock()/unlock()`.
  - `node_group_scale_lock_duration` metric now a historgram with fib seconds as buckets up to 55 then +Inf via default.

* escalator/60:
  - Added metric `escalator_node_group_scale_lock_check_was_locked` to count how many time the lock was probed and found held/locked.

* escalator/60: Added dependency recalculation and wired to dependency retrieval if sources have changed imports via Makefile.

* escalator/60: Switched `node_group_scale_lock_duration` to a Summary vector for metrics.

* escalator/60: Added new metrics docs.

* escalator/60: Reverted scale lock duration metric type to a histogram with 60 second buckets from 1 … 30.

Merge pull request atlassian#128 from atlassian/fix-metrics-for-nonle…

…ader

Fixes atlassian#129

Merge pull request atlassian#123 from atlassian/aprice/k8s_tests

k8s tests

add support for auto-discovering node group options from the cloud pr…

…ovider (atlassian#124)

This adds support for auto-discovering min_nodes and max_nodes from the cloud provider
for each node group. The auto-discovering/setting of the values is only done on the
initial start of Escalator. It can be enabled/disabled by setting both min_nodes and
max_nodes to 0 or removing them from the node groups config.