diff --git a/cluster/juju/layers/kubernetes-master/README.md b/cluster/juju/layers/kubernetes-master/README.md index c1cc84a6cf0..6279a54cfc9 100644 --- a/cluster/juju/layers/kubernetes-master/README.md +++ b/cluster/juju/layers/kubernetes-master/README.md @@ -1,19 +1,19 @@ # Kubernetes-master -[Kubernetes](http://kubernetes.io/) is an open source system for managing +[Kubernetes](http://kubernetes.io/) is an open source system for managing application containers across a cluster of hosts. The Kubernetes project was -started by Google in 2014, combining the experience of running production +started by Google in 2014, combining the experience of running production workloads combined with best practices from the community. The Kubernetes project defines some new terms that may be unfamiliar to users -or operators. For more information please refer to the concept guide in the +or operators. For more information please refer to the concept guide in the [getting started guide](https://kubernetes.io/docs/home/). -This charm is an encapsulation of the Kubernetes master processes and the +This charm is an encapsulation of the Kubernetes master processes and the operations to run on any cloud for the entire lifecycle of the cluster. This charm is built from other charm layers using the Juju reactive framework. -The other layers focus on specific subset of operations making this layer +The other layers focus on specific subset of operations making this layer specific to operations of Kubernetes master processes. # Deployment @@ -23,15 +23,15 @@ charms to model a complete Kubernetes cluster. A Kubernetes cluster needs a distributed key value store such as [Etcd](https://coreos.com/etcd/) and the kubernetes-worker charm which delivers the Kubernetes node services. A cluster requires a Software Defined Network (SDN) and Transport Layer Security (TLS) so -the components in a cluster communicate securely. +the components in a cluster communicate securely. -Please take a look at the [Canonical Distribution of Kubernetes](https://jujucharms.com/canonical-kubernetes/) -or the [Kubernetes core](https://jujucharms.com/kubernetes-core/) bundles for +Please take a look at the [Canonical Distribution of Kubernetes](https://jujucharms.com/canonical-kubernetes/) +or the [Kubernetes core](https://jujucharms.com/kubernetes-core/) bundles for examples of complete models of Kubernetes clusters. # Resources -The kubernetes-master charm takes advantage of the [Juju Resources](https://jujucharms.com/docs/2.0/developer-resources) +The kubernetes-master charm takes advantage of the [Juju Resources](https://jujucharms.com/docs/2.0/developer-resources) feature to deliver the Kubernetes software. In deployments on public clouds the Charm Store provides the resource to the @@ -40,9 +40,41 @@ firewall rules may not be able to contact the Charm Store. In these network restricted environments the resource can be uploaded to the model by the Juju operator. +#### Snap Refresh + +The kubernetes resources used by this charm are snap packages. When not +specified during deployment, these resources come from the public store. By +default, the `snapd` daemon will refresh all snaps installed from the store +four (4) times per day. A charm configuration option is provided for operators +to control this refresh frequency. + +>NOTE: this is a global configuration option and will affect the refresh +time for all snaps installed on a system. + +Examples: + +```sh +## refresh kubernetes-master snaps every tuesday +juju config kubernetes-master snapd_refresh="tue" + +## refresh snaps at 11pm on the last (5th) friday of the month +juju config kubernetes-master snapd_refresh="fri5,23:00" + +## delay the refresh as long as possible +juju config kubernetes-master snapd_refresh="max" + +## use the system default refresh timer +juju config kubernetes-master snapd_refresh="" +``` + +For more information on the possible values for `snapd_refresh`, see the +*refresh.timer* section in the [system options][] documentation. + +[system options]: https://forum.snapcraft.io/t/system-options/87 + # Configuration -This charm supports some configuration options to set up a Kubernetes cluster +This charm supports some configuration options to set up a Kubernetes cluster that works in your environment: #### dns_domain @@ -61,14 +93,14 @@ Enable RBAC and Node authorisation. # DNS for the cluster The DNS add-on allows the pods to have a DNS names in addition to IP addresses. -The Kubernetes cluster DNS server (based off the SkyDNS library) supports -forward lookups (A records), service lookups (SRV records) and reverse IP +The Kubernetes cluster DNS server (based off the SkyDNS library) supports +forward lookups (A records), service lookups (SRV records) and reverse IP address lookups (PTR records). More information about the DNS can be obtained from the [Kubernetes DNS admin guide](http://kubernetes.io/docs/admin/dns/). # Actions -The kubernetes-master charm models a few one time operations called +The kubernetes-master charm models a few one time operations called [Juju actions](https://jujucharms.com/docs/stable/actions) that can be run by Juju users. @@ -80,7 +112,7 @@ requires a relation to the ceph-mon charm before it can create the volume. #### restart -This action restarts the master processes `kube-apiserver`, +This action restarts the master processes `kube-apiserver`, `kube-controller-manager`, and `kube-scheduler` when the user needs a restart. # More information @@ -93,7 +125,7 @@ This action restarts the master processes `kube-apiserver`, # Contact The kubernetes-master charm is free and open source operations created -by the containers team at Canonical. +by the containers team at Canonical. Canonical also offers enterprise support and customization services. Please refer to the [Kubernetes product page](https://www.ubuntu.com/cloud/kubernetes) diff --git a/cluster/juju/layers/kubernetes-master/config.yaml b/cluster/juju/layers/kubernetes-master/config.yaml index 01171c058fd..e6acd542526 100644 --- a/cluster/juju/layers/kubernetes-master/config.yaml +++ b/cluster/juju/layers/kubernetes-master/config.yaml @@ -147,6 +147,15 @@ options: default: true description: | If true the metrics server for Kubernetes will be deployed onto the cluster. + snapd_refresh: + default: "max" + type: string + description: | + How often snapd handles updates for installed snaps. Setting an empty + string will check 4x per day. Set to "max" to delay the refresh as long + as possible. You may also set a custom string as described in the + 'refresh.timer' section here: + https://forum.snapcraft.io/t/system-options/87 default-storage: type: string default: "auto" diff --git a/cluster/juju/layers/kubernetes-master/layer.yaml b/cluster/juju/layers/kubernetes-master/layer.yaml index 44488b5ab2b..b1a1ed22c1a 100644 --- a/cluster/juju/layers/kubernetes-master/layer.yaml +++ b/cluster/juju/layers/kubernetes-master/layer.yaml @@ -18,6 +18,8 @@ includes: - 'interface:aws-integration' - 'interface:gcp-integration' - 'interface:openstack-integration' + - 'interface:vsphere-integration' + - 'interface:azure-integration' options: basic: packages: diff --git a/cluster/juju/layers/kubernetes-master/metadata.yaml b/cluster/juju/layers/kubernetes-master/metadata.yaml index 8cbc0b50c94..03746ad95a2 100644 --- a/cluster/juju/layers/kubernetes-master/metadata.yaml +++ b/cluster/juju/layers/kubernetes-master/metadata.yaml @@ -47,6 +47,10 @@ requires: interface: gcp-integration openstack: interface: openstack-integration + vsphere: + interface: vsphere-integration + azure: + interface: azure-integration resources: kubectl: type: file diff --git a/cluster/juju/layers/kubernetes-master/reactive/kubernetes_master.py b/cluster/juju/layers/kubernetes-master/reactive/kubernetes_master.py index 8842930d588..6652aeafb12 100644 --- a/cluster/juju/layers/kubernetes-master/reactive/kubernetes_master.py +++ b/cluster/juju/layers/kubernetes-master/reactive/kubernetes_master.py @@ -106,6 +106,15 @@ def check_for_upgrade_needed(): '''An upgrade charm event was triggered by Juju, react to that here.''' hookenv.status_set('maintenance', 'Checking resources') + # migrate to new flags + if is_state('kubernetes-master.restarted-for-cloud'): + remove_state('kubernetes-master.restarted-for-cloud') + set_state('kubernetes-master.cloud.ready') + if is_state('kubernetes-master.cloud-request-sent'): + # minor change, just for consistency + remove_state('kubernetes-master.cloud-request-sent') + set_state('kubernetes-master.cloud.request-sent') + migrate_from_pre_snaps() add_rbac_roles() set_state('reconfigure.authentication.setup') @@ -427,6 +436,38 @@ def set_app_version(): hookenv.application_version_set(version.split(b' v')[-1].rstrip()) +@when('kubernetes-master.snaps.installed') +@when('snap.refresh.set') +@when('leadership.is_leader') +def process_snapd_timer(): + ''' Set the snapd refresh timer on the leader so all cluster members + (present and future) will refresh near the same time. ''' + # Get the current snapd refresh timer; we know layer-snap has set this + # when the 'snap.refresh.set' flag is present. + timer = snap.get(snapname='core', key='refresh.timer').decode('utf-8') + + # The first time through, data_changed will be true. Subsequent calls + # should only update leader data if something changed. + if data_changed('master_snapd_refresh', timer): + hookenv.log('setting snapd_refresh timer to: {}'.format(timer)) + leader_set({'snapd_refresh': timer}) + + +@when('kubernetes-master.snaps.installed') +@when('snap.refresh.set') +@when('leadership.changed.snapd_refresh') +@when_not('leadership.is_leader') +def set_snapd_timer(): + ''' Set the snapd refresh.timer on non-leader cluster members. ''' + # NB: This method should only be run when 'snap.refresh.set' is present. + # Layer-snap will always set a core refresh.timer, which may not be the + # same as our leader. Gating with 'snap.refresh.set' ensures layer-snap + # has finished and we are free to set our config to the leader's timer. + timer = leader_get('snapd_refresh') + hookenv.log('setting snapd_refresh timer to: {}'.format(timer)) + snap.set_refresh_timer(timer) + + @hookenv.atexit def set_final_status(): ''' Set the final status of the charm as we leave hook execution ''' @@ -435,6 +476,22 @@ def set_final_status(): except NotImplementedError: goal_state = {} + vsphere_joined = is_state('endpoint.vsphere.joined') + azure_joined = is_state('endpoint.azure.joined') + cloud_blocked = is_state('kubernetes-master.cloud.blocked') + if vsphere_joined and cloud_blocked: + hookenv.status_set('blocked', + 'vSphere integration requires K8s 1.12 or greater') + return + if azure_joined and cloud_blocked: + hookenv.status_set('blocked', + 'Azure integration requires K8s 1.11 or greater') + return + + if is_state('kubernetes-master.cloud.pending'): + hookenv.status_set('waiting', 'Waiting for cloud integration') + return + if not is_state('kube-api-endpoint.available'): if 'kube-api-endpoint' in goal_state.get('relations', {}): status = 'waiting' @@ -478,16 +535,6 @@ def set_final_status(): hookenv.status_set('waiting', 'Waiting to retry addon deployment') return - req_sent = is_state('kubernetes-master.cloud-request-sent') - openstack_joined = is_state('endpoint.openstack.joined') - cloud_req = req_sent or openstack_joined - aws_ready = is_state('endpoint.aws.ready') - gcp_ready = is_state('endpoint.gcp.ready') - openstack_ready = is_state('endpoint.openstack.ready') - cloud_ready = aws_ready or gcp_ready or openstack_ready - if cloud_req and not cloud_ready: - hookenv.status_set('waiting', 'waiting for cloud integration') - if addons_configured and not all_kube_system_pods_running(): hookenv.status_set('waiting', 'Waiting for kube-system pods to start') return @@ -525,7 +572,9 @@ def master_services_down(): @when('etcd.available', 'tls_client.server.certificate.saved', 'authentication.setup') @when('leadership.set.auto_storage_backend') -@when_not('kubernetes-master.components.started') +@when_not('kubernetes-master.components.started', + 'kubernetes-master.cloud.pending', + 'kubernetes-master.cloud.blocked') def start_master(etcd): '''Run the Kubernetes master components.''' hookenv.status_set('maintenance', @@ -1337,6 +1386,15 @@ def configure_apiserver(etcd_connection_string): cloud_config_path = _cloud_config_path('kube-apiserver') api_opts['cloud-provider'] = 'openstack' api_opts['cloud-config'] = str(cloud_config_path) + elif (is_state('endpoint.vsphere.ready') and + get_version('kube-apiserver') >= (1, 12)): + cloud_config_path = _cloud_config_path('kube-apiserver') + api_opts['cloud-provider'] = 'vsphere' + api_opts['cloud-config'] = str(cloud_config_path) + elif is_state('endpoint.azure.ready'): + cloud_config_path = _cloud_config_path('kube-apiserver') + api_opts['cloud-provider'] = 'azure' + api_opts['cloud-config'] = str(cloud_config_path) audit_root = '/root/cdk/audit' os.makedirs(audit_root, exist_ok=True) @@ -1394,6 +1452,15 @@ def configure_controller_manager(): cloud_config_path = _cloud_config_path('kube-controller-manager') controller_opts['cloud-provider'] = 'openstack' controller_opts['cloud-config'] = str(cloud_config_path) + elif (is_state('endpoint.vsphere.ready') and + get_version('kube-apiserver') >= (1, 12)): + cloud_config_path = _cloud_config_path('kube-controller-manager') + controller_opts['cloud-provider'] = 'vsphere' + controller_opts['cloud-config'] = str(cloud_config_path) + elif is_state('endpoint.azure.ready'): + cloud_config_path = _cloud_config_path('kube-controller-manager') + controller_opts['cloud-provider'] = 'azure' + controller_opts['cloud-config'] = str(cloud_config_path) configure_kubernetes_service('kube-controller-manager', controller_opts, 'controller-manager-extra-args') @@ -1620,9 +1687,29 @@ def clear_cluster_tag_sent(): @when_any('endpoint.aws.joined', - 'endpoint.gcp.joined') + 'endpoint.gcp.joined', + 'endpoint.openstack.joined', + 'endpoint.vsphere.joined', + 'endpoint.azure.joined') +@when_not('kubernetes-master.cloud.ready') +def set_cloud_pending(): + k8s_version = get_version('kube-apiserver') + k8s_1_11 = k8s_version >= (1, 11) + k8s_1_12 = k8s_version >= (1, 12) + vsphere_joined = is_state('endpoint.vsphere.joined') + azure_joined = is_state('endpoint.azure.joined') + if (vsphere_joined and not k8s_1_12) or (azure_joined and not k8s_1_11): + set_state('kubernetes-master.cloud.blocked') + else: + remove_state('kubernetes-master.cloud.blocked') + set_state('kubernetes-master.cloud.pending') + + +@when_any('endpoint.aws.joined', + 'endpoint.gcp.joined', + 'endpoint.azure.joined') @when('leadership.set.cluster_tag') -@when_not('kubernetes-master.cloud-request-sent') +@when_not('kubernetes-master.cloud.request-sent') def request_integration(): hookenv.status_set('maintenance', 'requesting cloud integration') cluster_tag = leader_get('cluster_tag') @@ -1648,32 +1735,55 @@ def request_integration(): }) cloud.enable_object_storage_management() cloud.enable_security_management() + elif is_state('endpoint.azure.joined'): + cloud = endpoint_from_flag('endpoint.azure.joined') + cloud.tag_instance({ + 'k8s-io-cluster-name': cluster_tag, + 'k8s-io-role-master': 'master', + }) + cloud.enable_object_storage_management() + cloud.enable_security_management() cloud.enable_instance_inspection() cloud.enable_network_management() cloud.enable_dns_management() cloud.enable_block_storage_management() - set_state('kubernetes-master.cloud-request-sent') + set_state('kubernetes-master.cloud.request-sent') @when_none('endpoint.aws.joined', - 'endpoint.gcp.joined') -@when('kubernetes-master.cloud-request-sent') -def clear_requested_integration(): - remove_state('kubernetes-master.cloud-request-sent') + 'endpoint.gcp.joined', + 'endpoint.openstack.joined', + 'endpoint.vsphere.joined', + 'endpoint.azure.joined') +def clear_cloud_flags(): + remove_state('kubernetes-master.cloud.pending') + remove_state('kubernetes-master.cloud.request-sent') + remove_state('kubernetes-master.cloud.blocked') + remove_state('kubernetes-master.cloud.ready') @when_any('endpoint.aws.ready', 'endpoint.gcp.ready', - 'endpoint.openstack.ready') -@when_not('kubernetes-master.restarted-for-cloud') -def restart_for_cloud(): + 'endpoint.openstack.ready', + 'endpoint.vsphere.ready', + 'endpoint.azure.ready') +@when_not('kubernetes-master.cloud.blocked', + 'kubernetes-master.cloud.ready') +def cloud_ready(): if is_state('endpoint.gcp.ready'): _write_gcp_snap_config('kube-apiserver') _write_gcp_snap_config('kube-controller-manager') elif is_state('endpoint.openstack.ready'): _write_openstack_snap_config('kube-apiserver') _write_openstack_snap_config('kube-controller-manager') - set_state('kubernetes-master.restarted-for-cloud') + elif is_state('endpoint.vsphere.ready'): + _write_vsphere_snap_config('kube-apiserver') + _write_vsphere_snap_config('kube-controller-manager') + elif is_state('endpoint.azure.ready'): + _write_azure_snap_config('kube-apiserver') + _write_azure_snap_config('kube-controller-manager') + remove_state('kubernetes-master.cloud.pending') + set_state('kubernetes-master.cloud.ready') remove_state('kubernetes-master.components.started') # force restart @@ -1739,3 +1849,55 @@ def _write_openstack_snap_config(component): 'tenant-name = {}'.format(openstack.project_name), 'domain-name = {}'.format(openstack.user_domain_name), ])) + + +def _write_vsphere_snap_config(component): + # vsphere requires additional cloud config + vsphere = endpoint_from_flag('endpoint.vsphere.ready') + + # NB: vsphere provider will ask kube-apiserver and -controller-manager to + # find a uuid from sysfs unless a global config value is set. Our strict + # snaps cannot read sysfs, so let's do it in the charm. An invalid uuid is + # not fatal for storage, but it will muddy the logs; try to get it right. + uuid_file = '/sys/class/dmi/id/product_uuid' + try: + with open(uuid_file, 'r') as f: + uuid = f.read().strip() + except IOError as err: + hookenv.log("Unable to read UUID from sysfs: {}".format(err)) + uuid = 'UNKNOWN' + + cloud_config_path = _cloud_config_path(component) + cloud_config_path.write_text('\n'.join([ + '[Global]', + 'insecure-flag = true', + 'datacenters = "{}"'.format(vsphere.datacenter), + 'vm-uuid = "VMware-{}"'.format(uuid), + '[VirtualCenter "{}"]'.format(vsphere.vsphere_ip), + 'user = {}'.format(vsphere.user), + 'password = {}'.format(vsphere.password), + '[Workspace]', + 'server = {}'.format(vsphere.vsphere_ip), + 'datacenter = "{}"'.format(vsphere.datacenter), + 'default-datastore = "{}"'.format(vsphere.datastore), + 'folder = "kubernetes"', + 'resourcepool-path = ""', + '[Disk]', + 'scsicontrollertype = "pvscsi"', + ])) + + +def _write_azure_snap_config(component): + azure = endpoint_from_flag('endpoint.azure.ready') + cloud_config_path = _cloud_config_path(component) + cloud_config_path.write_text(json.dumps({ + 'useInstanceMetadata': True, + 'useManagedIdentityExtension': True, + 'subscriptionId': azure.subscription_id, + 'resourceGroup': azure.resource_group, + 'location': azure.resource_group_location, + 'vnetName': azure.vnet_name, + 'vnetResourceGroup': azure.vnet_resource_group, + 'subnetName': azure.subnet_name, + 'securityGroupName': azure.security_group_name, + })) diff --git a/cluster/juju/layers/kubernetes-worker/README.md b/cluster/juju/layers/kubernetes-worker/README.md index 965199eaf75..a661696f653 100644 --- a/cluster/juju/layers/kubernetes-worker/README.md +++ b/cluster/juju/layers/kubernetes-worker/README.md @@ -27,6 +27,38 @@ To add additional compute capacity to your Kubernetes workers, you may join any related kubernetes-master, and enlist themselves as ready once the deployment is complete. +## Snap Configuration + +The kubernetes resources used by this charm are snap packages. When not +specified during deployment, these resources come from the public store. By +default, the `snapd` daemon will refresh all snaps installed from the store +four (4) times per day. A charm configuration option is provided for operators +to control this refresh frequency. + +>NOTE: this is a global configuration option and will affect the refresh +time for all snaps installed on a system. + +Examples: + +```sh +## refresh kubernetes-worker snaps every tuesday +juju config kubernetes-worker snapd_refresh="tue" + +## refresh snaps at 11pm on the last (5th) friday of the month +juju config kubernetes-worker snapd_refresh="fri5,23:00" + +## delay the refresh as long as possible +juju config kubernetes-worker snapd_refresh="max" + +## use the system default refresh timer +juju config kubernetes-worker snapd_refresh="" +``` + +For more information on the possible values for `snapd_refresh`, see the +*refresh.timer* section in the [system options][] documentation. + +[system options]: https://forum.snapcraft.io/t/system-options/87 + ## Operational actions The kubernetes-worker charm supports the following Operational Actions: @@ -89,7 +121,7 @@ service is not reachable. Note: When debugging connection issues with NodePort services, its important to first check the kube-proxy service on the worker units. If kube-proxy is not running, the associated port-mapping will not be configured in the iptables -rulechains. +rulechains. If you need to close the NodePort once a workload has been terminated, you can follow the same steps inversely. @@ -97,4 +129,3 @@ follow the same steps inversely. ``` juju run --application kubernetes-worker close-port 30510 ``` - diff --git a/cluster/juju/layers/kubernetes-worker/config.yaml b/cluster/juju/layers/kubernetes-worker/config.yaml index 7123fc21b7b..2f15134f259 100644 --- a/cluster/juju/layers/kubernetes-worker/config.yaml +++ b/cluster/juju/layers/kubernetes-worker/config.yaml @@ -80,3 +80,29 @@ options: description: | Docker image to use for the default backend. Auto will select an image based on architecture. + snapd_refresh: + default: "max" + type: string + description: | + How often snapd handles updates for installed snaps. Setting an empty + string will check 4x per day. Set to "max" to delay the refresh as long + as possible. You may also set a custom string as described in the + 'refresh.timer' section here: + https://forum.snapcraft.io/t/system-options/87 + kubelet-extra-config: + default: "{}" + type: string + description: | + Extra configuration to be passed to kubelet. Any values specified in this + config will be merged into a KubeletConfiguration file that is passed to + the kubelet service via the --config flag. This can be used to override + values provided by the charm. + + Requires Kubernetes 1.10+. + + The value for this config must be a YAML mapping that can be safely + merged with a KubeletConfiguration file. For example: + {evictionHard: {memory.available: 200Mi}} + + For more information about KubeletConfiguration, see upstream docs: + https://kubernetes.io/docs/tasks/administer-cluster/kubelet-config-file/ diff --git a/cluster/juju/layers/kubernetes-worker/layer.yaml b/cluster/juju/layers/kubernetes-worker/layer.yaml index af247e80f0a..e7958853ae5 100644 --- a/cluster/juju/layers/kubernetes-worker/layer.yaml +++ b/cluster/juju/layers/kubernetes-worker/layer.yaml @@ -3,6 +3,7 @@ includes: - 'layer:basic' - 'layer:debug' - 'layer:snap' + - 'layer:leadership' - 'layer:docker' - 'layer:metrics' - 'layer:nagios' @@ -15,6 +16,8 @@ includes: - 'interface:aws-integration' - 'interface:gcp-integration' - 'interface:openstack-integration' + - 'interface:vsphere-integration' + - 'interface:azure-integration' - 'interface:mount' config: deletes: diff --git a/cluster/juju/layers/kubernetes-worker/metadata.yaml b/cluster/juju/layers/kubernetes-worker/metadata.yaml index 3f3a83c47fb..d4040fb449b 100644 --- a/cluster/juju/layers/kubernetes-worker/metadata.yaml +++ b/cluster/juju/layers/kubernetes-worker/metadata.yaml @@ -36,6 +36,10 @@ requires: interface: gcp-integration openstack: interface: openstack-integration + vsphere: + interface: vsphere-integration + azure: + interface: azure-integration nfs: interface: mount provides: diff --git a/cluster/juju/layers/kubernetes-worker/reactive/kubernetes_worker.py b/cluster/juju/layers/kubernetes-worker/reactive/kubernetes_worker.py index 215fa7a3fea..52d4a4d0eb6 100644 --- a/cluster/juju/layers/kubernetes-worker/reactive/kubernetes_worker.py +++ b/cluster/juju/layers/kubernetes-worker/reactive/kubernetes_worker.py @@ -21,6 +21,9 @@ import random import shutil import subprocess import time +import yaml + +from charms.leadership import leader_get, leader_set from pathlib import Path from shlex import split @@ -61,6 +64,15 @@ db = unitdata.kv() @hook('upgrade-charm') def upgrade_charm(): + # migrate to new flags + if is_state('kubernetes-worker.restarted-for-cloud'): + remove_state('kubernetes-worker.restarted-for-cloud') + set_state('kubernetes-worker.cloud.ready') + if is_state('kubernetes-worker.cloud-request-sent'): + # minor change, just for consistency + remove_state('kubernetes-worker.cloud-request-sent') + set_state('kubernetes-worker.cloud.request-sent') + # Trigger removal of PPA docker installation if it was previously set. set_state('config.changed.install_from_upstream') hookenv.atexit(remove_state, 'config.changed.install_from_upstream') @@ -189,13 +201,6 @@ def channel_changed(): set_upgrade_needed() -@when('kubernetes-worker.snaps.upgrade-needed') -@when_not('kubernetes-worker.snaps.upgrade-specified') -def upgrade_needed_status(): - msg = 'Needs manual upgrade, run the upgrade action' - hookenv.status_set('blocked', msg) - - @when('kubernetes-worker.snaps.upgrade-specified') def install_snaps(): channel = hookenv.config('channel') @@ -290,26 +295,73 @@ def set_app_version(): @when('kubernetes-worker.snaps.installed') -@when_not('kube-control.dns.available') -def notify_user_transient_status(): - ''' Notify to the user we are in a transient state and the application - is still converging. Potentially remotely, or we may be in a detached loop - wait state ''' +@when('snap.refresh.set') +@when('leadership.is_leader') +def process_snapd_timer(): + ''' Set the snapd refresh timer on the leader so all cluster members + (present and future) will refresh near the same time. ''' + # Get the current snapd refresh timer; we know layer-snap has set this + # when the 'snap.refresh.set' flag is present. + timer = snap.get(snapname='core', key='refresh.timer').decode('utf-8') - # During deployment the worker has to start kubelet without cluster dns - # configured. If this is the first unit online in a service pool waiting - # to self host the dns pod, and configure itself to query the dns service - # declared in the kube-system namespace - - hookenv.status_set('waiting', 'Waiting for cluster DNS.') + # The first time through, data_changed will be true. Subsequent calls + # should only update leader data if something changed. + if data_changed('worker_snapd_refresh', timer): + hookenv.log('setting snapd_refresh timer to: {}'.format(timer)) + leader_set({'snapd_refresh': timer}) -@when('kubernetes-worker.snaps.installed', - 'kube-control.dns.available') -@when_not('kubernetes-worker.snaps.upgrade-needed') -def charm_status(kube_control): +@when('kubernetes-worker.snaps.installed') +@when('snap.refresh.set') +@when('leadership.changed.snapd_refresh') +@when_not('leadership.is_leader') +def set_snapd_timer(): + ''' Set the snapd refresh.timer on non-leader cluster members. ''' + # NB: This method should only be run when 'snap.refresh.set' is present. + # Layer-snap will always set a core refresh.timer, which may not be the + # same as our leader. Gating with 'snap.refresh.set' ensures layer-snap + # has finished and we are free to set our config to the leader's timer. + timer = leader_get('snapd_refresh') + hookenv.log('setting snapd_refresh timer to: {}'.format(timer)) + snap.set_refresh_timer(timer) + + +@hookenv.atexit +def charm_status(): '''Update the status message with the current status of kubelet.''' - update_kubelet_status() + vsphere_joined = is_state('endpoint.vsphere.joined') + azure_joined = is_state('endpoint.azure.joined') + cloud_blocked = is_state('kubernetes-worker.cloud.blocked') + if vsphere_joined and cloud_blocked: + hookenv.status_set('blocked', + 'vSphere integration requires K8s 1.12 or greater') + return + if azure_joined and cloud_blocked: + hookenv.status_set('blocked', + 'Azure integration requires K8s 1.11 or greater') + return + if is_state('kubernetes-worker.cloud.pending'): + hookenv.status_set('waiting', 'Waiting for cloud integration') + return + if not is_state('kube-control.dns.available'): + # During deployment the worker has to start kubelet without cluster dns + # configured. If this is the first unit online in a service pool + # waiting to self host the dns pod, and configure itself to query the + # dns service declared in the kube-system namespace + hookenv.status_set('waiting', 'Waiting for cluster DNS.') + return + if is_state('kubernetes-worker.snaps.upgrade-specified'): + hookenv.status_set('waiting', 'Upgrade pending') + return + if is_state('kubernetes-worker.snaps.upgrade-needed'): + hookenv.status_set('blocked', + 'Needs manual upgrade, run the upgrade action') + return + if is_state('kubernetes-worker.snaps.installed'): + update_kubelet_status() + return + else: + pass # will have been set by snap layer or other handler def update_kubelet_status(): @@ -394,6 +446,8 @@ def watch_for_changes(kube_api, kube_control, cni): 'kube-control.dns.available', 'kube-control.auth.available', 'cni.available', 'kubernetes-worker.restart-needed', 'worker.auth.bootstrapped') +@when_not('kubernetes-worker.cloud.pending', + 'kubernetes-worker.cloud.blocked') def start_worker(kube_api, kube_control, auth_control, cni): ''' Start kubelet using the provided API and DNS info.''' servers = get_kube_api_servers(kube_api) @@ -517,8 +571,9 @@ def apply_node_labels(): @when_any('config.changed.kubelet-extra-args', - 'config.changed.proxy-extra-args') -def extra_args_changed(): + 'config.changed.proxy-extra-args', + 'config.changed.kubelet-extra-config') +def config_changed_requires_restart(): set_state('kubernetes-worker.restart-needed') @@ -639,6 +694,20 @@ def configure_kubernetes_service(service, base_args, extra_args_key): db.set(prev_args_key, args) +def merge_kubelet_extra_config(config, extra_config): + ''' Updates config to include the contents of extra_config. This is done + recursively to allow deeply nested dictionaries to be merged. + + This is destructive: it modifies the config dict that is passed in. + ''' + for k, extra_config_value in extra_config.items(): + if isinstance(extra_config_value, dict): + config_value = config.setdefault(k, {}) + merge_kubelet_extra_config(config_value, extra_config_value) + else: + config[k] = extra_config_value + + def configure_kubelet(dns, ingress_ip): layer_options = layer.options('tls-client') ca_cert_path = layer_options.get('ca_certificate_path') @@ -650,29 +719,10 @@ def configure_kubelet(dns, ingress_ip): kubelet_opts['kubeconfig'] = kubeconfig_path kubelet_opts['network-plugin'] = 'cni' kubelet_opts['v'] = '0' - kubelet_opts['address'] = '0.0.0.0' - kubelet_opts['port'] = '10250' - kubelet_opts['cluster-domain'] = dns['domain'] - kubelet_opts['anonymous-auth'] = 'false' - kubelet_opts['client-ca-file'] = ca_cert_path - kubelet_opts['tls-cert-file'] = server_cert_path - kubelet_opts['tls-private-key-file'] = server_key_path kubelet_opts['logtostderr'] = 'true' - kubelet_opts['fail-swap-on'] = 'false' kubelet_opts['node-ip'] = ingress_ip - - if (dns['enable-kube-dns']): - kubelet_opts['cluster-dns'] = dns['sdn-ip'] - - # set --allow-privileged flag for kubelet kubelet_opts['allow-privileged'] = set_privileged() - if is_state('kubernetes-worker.gpu.enabled'): - hookenv.log('Adding ' - '--feature-gates=DevicePlugins=true ' - 'to kubelet') - kubelet_opts['feature-gates'] = 'DevicePlugins=true' - if is_state('endpoint.aws.ready'): kubelet_opts['cloud-provider'] = 'aws' elif is_state('endpoint.gcp.ready'): @@ -683,6 +733,79 @@ def configure_kubelet(dns, ingress_ip): cloud_config_path = _cloud_config_path('kubelet') kubelet_opts['cloud-provider'] = 'openstack' kubelet_opts['cloud-config'] = str(cloud_config_path) + elif is_state('endpoint.vsphere.joined'): + # vsphere just needs to be joined on the worker (vs 'ready') + cloud_config_path = _cloud_config_path('kubelet') + kubelet_opts['cloud-provider'] = 'vsphere' + # NB: vsphere maps node product-id to its uuid (no config file needed). + uuid_file = '/sys/class/dmi/id/product_uuid' + with open(uuid_file, 'r') as f: + uuid = f.read().strip() + kubelet_opts['provider-id'] = 'vsphere://{}'.format(uuid) + elif is_state('endpoint.azure.ready'): + azure = endpoint_from_flag('endpoint.azure.ready') + cloud_config_path = _cloud_config_path('kubelet') + kubelet_opts['cloud-provider'] = 'azure' + kubelet_opts['cloud-config'] = str(cloud_config_path) + kubelet_opts['provider-id'] = azure.vm_id + + if get_version('kubelet') >= (1, 10): + # Put together the KubeletConfiguration data + kubelet_config = { + 'apiVersion': 'kubelet.config.k8s.io/v1beta1', + 'kind': 'KubeletConfiguration', + 'address': '0.0.0.0', + 'authentication': { + 'anonymous': { + 'enabled': False + }, + 'x509': { + 'clientCAFile': ca_cert_path + } + }, + 'clusterDomain': dns['domain'], + 'failSwapOn': False, + 'port': 10250, + 'tlsCertFile': server_cert_path, + 'tlsPrivateKeyFile': server_key_path + } + if dns['enable-kube-dns']: + kubelet_config['clusterDNS'] = [dns['sdn-ip']] + if is_state('kubernetes-worker.gpu.enabled'): + kubelet_config['featureGates'] = { + 'DevicePlugins': True + } + + # Add kubelet-extra-config. This needs to happen last so that it + # overrides any config provided by the charm. + kubelet_extra_config = hookenv.config('kubelet-extra-config') + kubelet_extra_config = yaml.load(kubelet_extra_config) + merge_kubelet_extra_config(kubelet_config, kubelet_extra_config) + + # Render the file and configure Kubelet to use it + os.makedirs('/root/cdk/kubelet', exist_ok=True) + with open('/root/cdk/kubelet/config.yaml', 'w') as f: + f.write('# Generated by kubernetes-worker charm, do not edit\n') + yaml.dump(kubelet_config, f) + kubelet_opts['config'] = '/root/cdk/kubelet/config.yaml' + else: + # NOTE: This is for 1.9. Once we've dropped 1.9 support, we can remove + # this whole block and the parent if statement. + kubelet_opts['address'] = '0.0.0.0' + kubelet_opts['anonymous-auth'] = 'false' + kubelet_opts['client-ca-file'] = ca_cert_path + kubelet_opts['cluster-domain'] = dns['domain'] + kubelet_opts['fail-swap-on'] = 'false' + kubelet_opts['port'] = '10250' + kubelet_opts['tls-cert-file'] = server_cert_path + kubelet_opts['tls-private-key-file'] = server_key_path + if dns['enable-kube-dns']: + kubelet_opts['cluster-dns'] = dns['sdn-ip'] + if is_state('kubernetes-worker.gpu.enabled'): + kubelet_opts['feature-gates'] = 'DevicePlugins=true' + + if get_version('kubelet') >= (1, 11): + kubelet_opts['dynamic-config-dir'] = '/root/cdk/kubelet/dynamic-config' configure_kubernetes_service('kubelet', kubelet_opts, 'kubelet-extra-args') @@ -794,10 +917,10 @@ def launch_default_ingress_controller(): 'ingress-ssl-chain-completion') context['ingress_image'] = config.get('nginx-image') if context['ingress_image'] == "" or context['ingress_image'] == "auto": - images = {'amd64': 'quay.io/kubernetes-ingress-controller/nginx-ingress-controller:0.16.1', # noqa - 'arm64': 'quay.io/kubernetes-ingress-controller/nginx-ingress-controller-arm64:0.16.1', # noqa - 's390x': 'quay.io/kubernetes-ingress-controller/nginx-ingress-controller-s390x:0.16.1', # noqa - 'ppc64el': 'quay.io/kubernetes-ingress-controller/nginx-ingress-controller-ppc64le:0.16.1', # noqa + images = {'amd64': 'quay.io/kubernetes-ingress-controller/nginx-ingress-controller:0.16.1', # noqa + 'arm64': 'quay.io/kubernetes-ingress-controller/nginx-ingress-controller-arm64:0.16.1', # noqa + 's390x': 'quay.io/kubernetes-ingress-controller/nginx-ingress-controller-s390x:0.16.1', # noqa + 'ppc64el': 'quay.io/kubernetes-ingress-controller/nginx-ingress-controller-ppc64le:0.16.1', # noqa } context['ingress_image'] = images.get(context['arch'], images['amd64']) if get_version('kubelet') < (1, 9): @@ -1104,6 +1227,10 @@ def get_node_name(): cloud_provider = 'gce' elif is_state('endpoint.openstack.ready'): cloud_provider = 'openstack' + elif is_state('endpoint.vsphere.ready'): + cloud_provider = 'vsphere' + elif is_state('endpoint.azure.ready'): + cloud_provider = 'azure' if cloud_provider == 'aws': return getfqdn().lower() else: @@ -1147,9 +1274,29 @@ def remove_label(label): @when_any('endpoint.aws.joined', - 'endpoint.gcp.joined') + 'endpoint.gcp.joined', + 'endpoint.openstack.joined', + 'endpoint.vsphere.joined', + 'endpoint.azure.joined') +@when_not('kubernetes-worker.cloud.ready') +def set_cloud_pending(): + k8s_version = get_version('kubelet') + k8s_1_11 = k8s_version >= (1, 11) + k8s_1_12 = k8s_version >= (1, 12) + vsphere_joined = is_state('endpoint.vsphere.joined') + azure_joined = is_state('endpoint.azure.joined') + if (vsphere_joined and not k8s_1_12) or (azure_joined and not k8s_1_11): + set_state('kubernetes-worker.cloud.blocked') + else: + remove_state('kubernetes-worker.cloud.blocked') + set_state('kubernetes-worker.cloud.pending') + + +@when_any('endpoint.aws.joined', + 'endpoint.gcp.joined', + 'endpoint.azure.joined') @when('kube-control.cluster_tag.available') -@when_not('kubernetes-worker.cloud-request-sent') +@when_not('kubernetes-worker.cloud.request-sent') def request_integration(): hookenv.status_set('maintenance', 'requesting cloud integration') kube_control = endpoint_from_flag('kube-control.cluster_tag.available') @@ -1172,29 +1319,47 @@ def request_integration(): 'k8s-io-cluster-name': cluster_tag, }) cloud.enable_object_storage_management() + elif is_state('endpoint.azure.joined'): + cloud = endpoint_from_flag('endpoint.azure.joined') + cloud.tag_instance({ + 'k8s-io-cluster-name': cluster_tag, + }) + cloud.enable_object_storage_management() cloud.enable_instance_inspection() cloud.enable_dns_management() - set_state('kubernetes-worker.cloud-request-sent') - hookenv.status_set('waiting', 'waiting for cloud integration') + set_state('kubernetes-worker.cloud.request-sent') + hookenv.status_set('waiting', 'Waiting for cloud integration') @when_none('endpoint.aws.joined', - 'endpoint.gcp.joined') -def clear_requested_integration(): - remove_state('kubernetes-worker.cloud-request-sent') + 'endpoint.gcp.joined', + 'endpoint.openstack.joined', + 'endpoint.vsphere.joined', + 'endpoint.azure.joined') +def clear_cloud_flags(): + remove_state('kubernetes-worker.cloud.pending') + remove_state('kubernetes-worker.cloud.request-sent') + remove_state('kubernetes-worker.cloud.blocked') + remove_state('kubernetes-worker.cloud.ready') @when_any('endpoint.aws.ready', 'endpoint.gcp.ready', - 'endpoint.openstack.ready') -@when_not('kubernetes-worker.restarted-for-cloud') -def restart_for_cloud(): + 'endpoint.openstack.ready', + 'endpoint.vsphere.ready', + 'endpoint.azure.ready') +@when_not('kubernetes-worker.cloud.blocked', + 'kubernetes-worker.cloud.ready') +def cloud_ready(): + remove_state('kubernetes-worker.cloud.pending') if is_state('endpoint.gcp.ready'): _write_gcp_snap_config('kubelet') elif is_state('endpoint.openstack.ready'): _write_openstack_snap_config('kubelet') - set_state('kubernetes-worker.restarted-for-cloud') - set_state('kubernetes-worker.restart-needed') + elif is_state('endpoint.azure.ready'): + _write_azure_snap_config('kubelet') + set_state('kubernetes-worker.cloud.ready') + set_state('kubernetes-worker.restart-needed') # force restart def _snap_common_path(component): @@ -1257,6 +1422,22 @@ def _write_openstack_snap_config(component): ])) +def _write_azure_snap_config(component): + azure = endpoint_from_flag('endpoint.azure.ready') + cloud_config_path = _cloud_config_path(component) + cloud_config_path.write_text(json.dumps({ + 'useInstanceMetadata': True, + 'useManagedIdentityExtension': True, + 'subscriptionId': azure.subscription_id, + 'resourceGroup': azure.resource_group, + 'location': azure.resource_group_location, + 'vnetName': azure.vnet_name, + 'vnetResourceGroup': azure.vnet_resource_group, + 'subnetName': azure.subnet_name, + 'securityGroupName': azure.security_group_name, + })) + + def get_first_mount(mount_relation): mount_relation_list = mount_relation.mounts() if mount_relation_list and len(mount_relation_list) > 0: