Releases: atlassian/escalator
v1.4.0
v1.3.0
This release contains a new metric for tracking the time taken for a new node to be created by the cloud provider and when it has been created in the Kubernetes API.
The metric requires an update to the IAM permissions for AWS so that Escalator can describe instances. See https://github.com/atlassian/escalator/tree/master/docs/deployment/aws#permissions for the updated AWS IAM policy.
Features
- Added histogram of how long new nodes take to register… #137
Docker Image
atlassian/escalator:v1.3.0
v1.2.1
v1.2.0
This release contains new metrics for monitoring the scale lock functionality and 2 bug fixes. The new metrics documentation can be found here - https://github.com/atlassian/escalator/blob/master/docs/metrics.md
Features
- Add scale lock summary metric to observe lock durations & count #134
Bug Fixes
- Simplify cloud provider node group access within controller #130
- Scale up to ceiling of ASG max size if scale up delta exceeds ASG max size #132
Docker Image
atlassian/escalator:v1.2.0
v1.1.1
v1.1.0
This release adds leader election via the --leader-elect flag, along with other options.
The leader election uses a ConfigMap for storing the lock, and election events can be viewed using describe configmap <configmapname>.
Config changes for Leader election
If this flag is enabled, the Pod that Escalator runs in will need an environment variable POD_NAME set, exposed via the Downward API, to have the leader annotations and the events have the name of the pod. If not, a UUID will be used instead. Check the Deployment or AWS Deployment sample configs for details. Escalator's ClusterRole will also need updates, see Escalator RBAC for the details.
New command-line parameters
The new parameters: (from Command line options)
--leader-elect
Enable leader election behaviour. Note that Escalator uses a ConfigMap for the leader lock, not an Endpoint.
--leader-elect-lease-duration
Sets how long a nonleader will wait before it attempts to require the leadership. Measured against time of last observed ack.
--leader-elect-renew-deadline
Sets how long an acting leader will retry refreshing leadership before giving up.
--leader-elect-retry-period
Sets how long all the clients will wait in between attempts of any action.
--leader-elect-config-namespace
Sets the namespace where the configmap used for locking will be created or looked for. Defaults to kube-system
--leader-elect-config-name
Sets the name of the configmap used for locking. Defaults to escalator-leader-elect
Changelog
- Added leader election, #56
- Added tests for k8s code.
- Updated deployment YAML files with required extra parameters.
Docker Image
atlassian/escalator:v1.1.0