Changes in MetalK8s

Release 1.0.0 (in development)

Breaking changes

PR #187 - no longer remove the MetalK8s 0.1.x Elasticsearch cluster upon upgrade (#160)

Features added

PR #191 - deploy PodDisruptionBudgets for Elasticsearch (#157)

PR #193 - update versions of kube-prometheus, Elasticsearch and Kubespray

PR #181 - format PersistentVolumes asynchronously (#173)

PR #201 - collect Calico metrics and deploy Grafana dashboards for them (#81)

PR #210 - deploy metrics-server using Helm (#146)

PR #189, PR #215 - collect nginx-ingress metrics and deploy a dashboard (#143)

PR #218 - update versions of Kibana and fluent-bit

Release 0.2.1 (in development)

Bugs fixed

#170 - rename ElasticSearch Example and Node Exporter Full Grafana dashboards (PR #188)

#196 - deploy the Elasticsearch Curator configuration we want to deploy instead of falling back to the chart default (PR #197)

Release 0.2.0

Note

Compatibility with future releases of MetalK8s is not guaranteed until version 1.0.0 is available. When deploying a cluster using pre-1.0 versions of this package, you may need to redeploy later.

Breaking changes

PR #159 - use upstream chart for Elasticsearch. Historical log data will be lost. Please see the pull-request description for manual steps required after upgrading a MetalK8s 0.1 cluster to MetalK8s 0.2 (#147)

PR #94 - flatten the storage configuration and allow more user defined storage related actions. Please see Upgrading from MetalK8s < 0.2.0 (#153)

Features added

PR #144 - update Kibana chart version

PR #145 - update the Cerebro chart, and pre-configure the MetalK8s Elasticsearch cluster

PR #154 - rework log collection architecture, now using Fluent Bit to capture logs, then forward to fluentd to aggregate them and batch-insert in Elasticsearch (#51)

PR #163 - update versions of Elasticsearch Exporter, nginx-ingress, kube-prometheus and Kubespray

Bugs fixed

PR #151 - fix debug clause var scoping

#150 - fix deployment of Elasticsearch, node and Prometheus Grafana dashboards (PR #158)

#139 - stabilize helm init (PR #167)

Known issues

#179 - some Grafana dashboard charts are not displaying any metrics

Release 0.1.1

Note

Compatibility with future releases of MetalK8s is not guaranteed until version 1.0.0 is available. When deploying a cluster using pre-1.0 versions of this package, you may need to redeploy later.

Features added

PR #11 - run the OpenStack ansible-hardening role on nodes to apply security hardening configurations from the Security Technical Implementation Guide (STIG) (#88)

PR #127 - deploy Cerebro to manage the Elasticsearch cluster (#126)

PR #138 - update versions of Fluentd, Kibana, Elasticsearch Exporter and Kubespray

PR #140 - set up kube-prometheus to monitor CoreDNS (cfr. PR #104)

Bugs fixed

#103 - set up host anti-affinity for Elasticsearch service scheduling (PR #113)

#120 - required facts not gathered when running the services playbook in isolation (PR #132)

PR #134 - fix bash-completion in the MetalK8s Docker image

Release 0.1.0

This marks the first release of MetalK8s.

Note

Compatibility with future releases of MetalK8s is not guaranteed until version 1.0.0 is available. When deploying a cluster using pre-1.0 versions of this package, you may need to redeploy later.

Incompatible changes

PR #106 - the Ansible playbook which used to be called metal-k8s.yml has been moved to playbooks/deploy.yml

Features added

PR #100 - disable Elasticsearch deployment by setting metalk8s_elasticsearch_enabled to false (#98)

PR #104 - kube-proxy now uses ipvs instead of iptables to route Service addresses, in preparation for Kubernetes 1.11. The ipvsadm tool is installed on all k8s-cluster hosts.

PR #104 - use CoreDNS instead of kubedns for in-cluster DNS services, in preparation for Kubernetes 1.11.

PR #113 - deploy the Prometheus node_exporter on k8s-cluster and etcd hosts instead of using a DaemonSet

Known issues

#62 - Elasticsearch Curator may not properly prune old logstash-* indices