From 81628f94bad4b303212bf77752f62c03728e0168 Mon Sep 17 00:00:00 2001 From: Jason DeTiberus Date: Tue, 17 Mar 2015 11:09:12 -0400 Subject: Fix hostname handling - always set hostname if hostname does not match openshift_hostname - Use local IP instead of public IP as hostname for workaround --- roles/openshift_common/README.md | 1 - roles/openshift_common/defaults/main.yml | 3 +-- roles/openshift_common/tasks/main.yml | 6 ++---- 3 files changed, 3 insertions(+), 7 deletions(-) diff --git a/roles/openshift_common/README.md b/roles/openshift_common/README.md index c2ae609ff..79076ed19 100644 --- a/roles/openshift_common/README.md +++ b/roles/openshift_common/README.md @@ -14,7 +14,6 @@ Role Variables | Name | Default value | | |-------------------------------|------------------------------|----------------------------------------| -| openshift_bind_ip | ansible_default_ipv4.address | IP to use for local binding | | openshift_debug_level | 0 | Global openshift debug log verbosity | | openshift_hostname_workaround | True | Workaround needed to set hostname to IP address | | openshift_hostname | openshift_public_ip if openshift_hostname_workaround else ansible_fqdn | hostname to use for this instance | diff --git a/roles/openshift_common/defaults/main.yml b/roles/openshift_common/defaults/main.yml index a541591fb..eb6edbc03 100644 --- a/roles/openshift_common/defaults/main.yml +++ b/roles/openshift_common/defaults/main.yml @@ -1,8 +1,7 @@ --- -openshift_bind_ip: "{{ ansible_default_ipv4.address }}" openshift_debug_level: 0 # TODO: Once openshift stops resolving hostnames for node queries remove # this... openshift_hostname_workaround: true -openshift_hostname: "{{ openshift_public_ip if openshift_hostname_workaround else ansible_fqdn }}" +openshift_hostname: "{{ ansible_default_ipv4.address if openshift_hostname_workaround else ansible_fqdn }}" diff --git a/roles/openshift_common/tasks/main.yml b/roles/openshift_common/tasks/main.yml index 728bba4e4..07737a71f 100644 --- a/roles/openshift_common/tasks/main.yml +++ b/roles/openshift_common/tasks/main.yml @@ -1,8 +1,6 @@ --- -# fixme: Once openshift stops resolving hostnames for node queries remove this... -- name: Set hostname to IP Addr (WORKAROUND) - hostname: name={{ openshift_bind_ip }} - when: openshift_hostname_workaround +- name: Set hostname + hostname: name={{ openshift_hostname }} - name: Configure local facts file file: path=/etc/ansible/facts.d/ state=directory mode=0750 -- cgit v1.2.3 From 7035459d20dd2d278b0a0e6ff96421639f6e0e34 Mon Sep 17 00:00:00 2001 From: Jason DeTiberus Date: Wed, 18 Mar 2015 00:05:51 -0400 Subject: Register node fixes - Set --hostname flag in node config in openshift_node role - Support some additional node attributes in openshift_node role - podCIDR - labels - annotations - Support both output types for openshift ex config view in openshift_register_node module - Support multiple api versions in openshift_register_node module - Support additional attributes in openshift_register_node module - annotations - labels - pod_cidr - external_ips (v1beta3, will be available after next kube rebase) - internal_ips (v1beta3, will be available after next kube rebase) - hostnames (v1beta3, will be available after next kube rebase) - external_id (v1beta3, will be available after next kube rebase) --- roles/openshift_node/defaults/main.yml | 8 +- .../library/openshift_register_node.py | 453 ++++++++++++++------- roles/openshift_node/tasks/main.yml | 14 +- 3 files changed, 333 insertions(+), 142 deletions(-) diff --git a/roles/openshift_node/defaults/main.yml b/roles/openshift_node/defaults/main.yml index c45524f16..e4d5ebfee 100644 --- a/roles/openshift_node/defaults/main.yml +++ b/roles/openshift_node/defaults/main.yml @@ -5,6 +5,8 @@ os_firewall_allow: - service: OpenShift kubelet port: 10250/tcp openshift_node_resources: - capacity: - cpu: - memory: + cpu: + memory: + cidr: +openshift_node_labels: {} +openshift_node_annotations: {} diff --git a/roles/openshift_node/library/openshift_register_node.py b/roles/openshift_node/library/openshift_register_node.py index 63079e59b..4922585d7 100644 --- a/roles/openshift_node/library/openshift_register_node.py +++ b/roles/openshift_node/library/openshift_register_node.py @@ -6,78 +6,315 @@ import os import multiprocessing import socket from subprocess import check_output, Popen +from decimal import * DOCUMENTATION = ''' --- -module: openshift_register_node -short_description: This module registers an openshift-node with an openshift-master -author: Jason DeTiberus -requirements: [ openshift-node ] -notes: Node resources can be specified using either the resources option or the following options: cpu, memory +module: kubernetes_register_node +short_description: Registers a kubernetes node with a master +description: + - Registers a kubernetes node with a master options: name: + default: null description: - - id for this node (usually the node fqdn) + - Identifier for this node (usually the node fqdn). required: true - hostIP: + api_verison: + choices: ['v1beta1', 'v1beta3'] + default: 'v1beta1' description: - - ip address for this node + - Kubernetes API version to use + required: true + host_ip: + default: null + description: + - IP Address to associate with the node when registering. + Available in the following API versions: v1beta1. required: false - cpu: + hostnames: + default: [] description: - - number of CPUs for this node + - Valid hostnames for this node. Available in the following API + versions: v1beta3. required: false - default: number of logical CPUs detected - memory: + external_ips: + default: [] description: - - Memory available for this node in bytes + - External IP Addresses for this node. Available in the following API + versions: v1beta3. required: false - default: 80% MemTotal - resources: + internal_ips: + default: [] description: - - A json string representing Node resources + - Internal IP Addresses for this node. Available in the following API + versions: v1beta3. + required: false + cpu: + default: null + description: + - Number of CPUs to allocate for this node. If not provided, then + the node will be registered to advertise the number of logical + CPUs available. When using the v1beta1 API, you must specify the + CPU count as a floating point number with no more than 3 decimal + places. API version v1beta3 and newer accepts arbitrary float + values. + required: false + memory: + default: null + description: + - Memory available for this node. If not provided, then the node + will be registered to advertise 80% of MemTotal as available + memory. When using the v1beta1 API, you must specify the memory + size in bytes. API version v1beta3 and newer accepts binary SI + and decimal SI values. required: false ''' EXAMPLES = ''' # Minimal node registration - openshift_register_node: name=ose3.node.example.com -# Node registration with all options (using cpu and memory options) +# Node registration using the v1beta1 API and assigning 1 CPU core and 10 GB of +# Memory - openshift_register_node: name: ose3.node.example.com + api_version: v1beta1 hostIP: 192.168.1.1 - apiVersion: v1beta1 cpu: 1 - memory: 1073741824 + memory: 500000000 -# Node registration with all options (using resources option) +# Node registration using the v1beta3 API, setting an alternate hostname, +# internalIP, externalIP and assigning 3.5 CPU cores and 1 TiB of Memory - openshift_register_node: name: ose3.node.example.com - hostIP: 192.168.1.1 - apiVersion: v1beta1 - resources: - capacity: - cpu: 1 - memory: 1073741824 + api_version: v1beta3 + external_ips: ['192.168.1.5'] + internal_ips: ['10.0.0.5'] + hostnames: ['ose2.node.internal.local'] + cpu: 3.5 + memory: 1Ti ''' + +class ClientConfigException(Exception): + pass + +class ClientConfig: + def __init__(self, client_opts, module): + _, output, error = module.run_command(["/usr/bin/openshift", "ex", + "config", "view", "-o", + "json"] + client_opts, + check_rc = True) + self.config = json.loads(output) + + if not (bool(self.config['clusters']) or + bool(self.config['contexts']) or + bool(self.config['current-context']) or + bool(self.config['users'])): + raise ClientConfigException(msg="Client config missing required " \ + "values", + output=output) + + def current_context(self): + return self.config['current-context'] + + def section_has_value(self, section_name, value): + section = self.config[section_name] + if isinstance(section, dict): + return value in section + else: + val = next((item for item in section + if item['name'] == value), None) + return val is not None + + def has_context(self, context): + return self.section_has_value('contexts', context) + + def has_user(self, user): + return self.section_has_value('users', user) + + def has_cluster(self, cluster): + return self.section_has_value('clusters', cluster) + + def get_value_for_context(self, context, attribute): + contexts = self.config['contexts'] + if isinstance(contexts, dict): + return contexts[context][attribute] + else: + return next((c['context'][attribute] for c in contexts + if c['name'] == context), None) + + def get_user_for_context(self, context): + return self.get_value_for_context(context, 'user') + + def get_cluster_for_context(self, context): + return self.get_value_for_context(context, 'cluster') + +class Util: + @staticmethod + def getLogicalCores(): + return multiprocessing.cpu_count() + + @staticmethod + def getMemoryPct(pct): + with open('/proc/meminfo', 'r') as mem: + for line in mem: + entries = line.split() + if str(entries.pop(0)) == 'MemTotal:': + mem_total_kb = Decimal(entries.pop(0)) + mem_capacity_kb = mem_total_kb * Decimal(pct) + return str(mem_capacity_kb.to_integral_value() * 1024) + + return "" + + @staticmethod + def remove_empty_elements(mapping): + if isinstance(mapping, dict): + m = mapping.copy() + for key, val in mapping.iteritems(): + if not val: + del m[key] + return m + else: + return mapping + +class NodeResources: + def __init__(self, version, cpu=None, memory=None): + if version == 'v1beta1': + self.resources = dict(capacity=dict()) + self.resources['capacity']['cpu'] = cpu if cpu else Util.getLogicalCores() + self.resources['capacity']['memory'] = memory if cpu else Util.getMemoryPct(.75) + + def get_resources(self): + return Util.remove_empty_elements(self.resources) + +class NodeSpec: + def __init__(self, version, cpu=None, memory=None, cidr=None, externalID=None): + if version == 'v1beta3': + self.spec = dict(podCIDR=cidr, externalID=externalID, + capacity=dict()) + self.spec['capacity']['cpu'] = cpu if cpu else Util.getLogicalCores() + self.spec['capacity']['memory'] = memory if memory else Util.getMemoryPct(.75) + + def get_spec(self): + return Util.remove_empty_elements(self.spec) + +class NodeStatus: + def addAddresses(self, addressType, addresses): + addressList = [] + for address in addresses: + addressList.append(dict(type=addressType, address=address)) + return addressList + + def __init__(self, version, externalIPs = [], internalIPs = [], + hostnames = []): + if version == 'v1beta3': + self.status = dict(addresses = addAddresses('ExternalIP', + externalIPs) + + addAddresses('InternalIP', + internalIPs) + + addAddresses('Hostname', + hostnames)) + + def get_status(self): + return Util.remove_empty_elements(self.status) + +class Node: + def __init__(self, module, client_opts, version='v1beta1', name=None, + hostIP = None, hostnames=[], externalIPs=[], internalIPs=[], + cpu=None, memory=None, labels=dict(), annotations=dict(), + podCIDR=None, externalID=None): + self.module = module + self.client_opts = client_opts + if version == 'v1beta1': + self.node = dict(id = name, + kind = 'Node', + apiVersion = version, + hostIP = hostIP, + resources = NodeResources(version, cpu, memory), + cidr = podCIDR, + labels = labels, + annotations = annotations + ) + elif version == 'v1beta3': + metadata = dict(name = name, + labels = labels, + annotations = annotations + ) + self.node = dict(kind = 'Node', + apiVersion = version, + metadata = metadata, + spec = NodeSpec(version, cpu, memory, podCIDR, + externalID), + status = NodeStatus(version, externalIPs, + internalIPs, hostnames), + ) + + def get_name(self): + if self.node['apiVersion'] == 'v1beta1': + return self.node['id'] + elif self.node['apiVersion'] == 'v1beta3': + return self.node['name'] + + def get_node(self): + node = self.node.copy() + if self.node['apiVersion'] == 'v1beta1': + node['resources'] = self.node['resources'].get_resources() + elif self.node['apiVersion'] == 'v1beta3': + node['spec'] = self.node['spec'].get_spec() + node['status'] = self.node['status'].get_status() + return Util.remove_empty_elements(node) + + def exists(self): + _, output, error = self.module.run_command(["/usr/bin/osc", "get", + "nodes"] + self.client_opts, + check_rc = True) + if re.search(self.module.params['name'], output, re.MULTILINE): + return True + return False + + def create(self): + cmd = ['/usr/bin/osc'] + self.client_opts + ['create', 'node', '-f', '-'] + rc, output, error = self.module.run_command(cmd, + data=self.module.jsonify(self.get_node())) + if rc != 0: + if re.search("minion \"%s\" already exists" % self.get_name(), + error): + self.module.exit_json(changed=False, + msg="node definition already exists", + node=self.get_node()) + else: + self.module.fail_json(msg="Node creation failed.", rc=rc, + output=output, error=error, + node=self.get_node()) + else: + return True + def main(): module = AnsibleModule( argument_spec = dict( - name = dict(required = True), - hostIP = dict(), - apiVersion = dict(), - cpu = dict(), - memory = dict(), - resources = dict(), - client_config = dict(), - client_cluster = dict(default = 'master'), - client_context = dict(default = 'master'), - client_user = dict(default = 'admin') + name = dict(required = True, type = 'str'), + host_ip = dict(type = 'str'), + hostnames = dict(type = 'list', default = []), + external_ips = dict(type = 'list', default = []), + internal_ips = dict(type = 'list', default = []), + api_version = dict(type = 'str', default = 'v1beta1', # TODO: after kube rebase, we can default to v1beta3 + choices = ['v1beta1', 'v1beta3']), + cpu = dict(type = 'str'), + memory = dict(type = 'str'), + labels = dict(type = 'dict', default = {}), # TODO: needs documented + annotations = dict(type = 'dict', default = {}), # TODO: needs documented + pod_cidr = dict(type = 'str'), # TODO: needs documented + external_id = dict(type = 'str'), # TODO: needs documented + client_config = dict(type = 'str'), # TODO: needs documented + client_cluster = dict(type = 'str', default = 'master'), # TODO: needs documented + client_context = dict(type = 'str', default = 'master'), # TODO: needs documented + client_user = dict(type = 'str', default = 'admin') # TODO: needs documented ), mutually_exclusive = [ - ['resources', 'cpu'], - ['resources', 'memory'] + ['host_ip', 'external_ips'], + ['host_ip', 'internal_ips'], + ['host_ip', 'hostnames'], ], supports_check_mode=True ) @@ -93,119 +330,61 @@ def main(): client_opts.append("--kubeconfig=%s" % module.params['client_config']) try: - output = check_output(["/usr/bin/openshift", "ex", "config", "view", - "-o", "json"] + client_opts, - stderr=subprocess.STDOUT) - except subprocess.CalledProcessError as e: - module.fail_json(msg="Failed to get client configuration", - command=e.cmd, returncode=e.returncode, output=e.output) - - config = json.loads(output) - if not (bool(config['clusters']) or bool(config['contexts']) or - bool(config['current-context']) or bool(config['users'])): - module.fail_json(msg="Client config missing required values", - output=output) + config = ClientConfig(client_opts, module) + except ClientConfigException as e: + module.fail_json(msg="Failed to get client configuration", exception=e) client_context = module.params['client_context'] - if client_context: - config_context = next((context for context in config['contexts'] - if context['name'] == client_context), None) - if not config_context: - module.fail_json(msg="Context %s not found in client config" % - client_context) - if not config['current-context'] or config['current-context'] != client_context: + if config.has_context(client_context): + if client_context != config.current_context(): client_opts.append("--context=%s" % client_context) + else: + module.fail_json(msg="Context %s not found in client config" % + client_context) client_user = module.params['client_user'] - if client_user: - config_user = next((user for user in config['users'] - if user['name'] == client_user), None) - if not config_user: - module.fail_json(msg="User %s not found in client config" % - client_user) - if client_user != config_context['context']['user']: + if config.has_user(client_user): + if client_user != config.get_user_for_context(client_context): client_opts.append("--user=%s" % client_user) + else: + module.fail_json(msg="User %s not found in client config" % + client_user) client_cluster = module.params['client_cluster'] - if client_cluster: - config_cluster = next((cluster for cluster in config['clusters'] - if cluster['name'] == client_cluster), None) - if not client_cluster: - module.fail_json(msg="Cluster %s not found in client config" % - client_cluster) - if client_cluster != config_context['context']['cluster']: + if config.has_cluster(client_cluster): + if client_cluster != config.get_cluster_for_context(client_cluster): client_opts.append("--cluster=%s" % client_cluster) + else: + module.fail_json(msg="Cluster %s not found in client config" % + client_cluster) - node_def = dict( - id = module.params['name'], - kind = 'Node', - apiVersion = 'v1beta1', - resources = dict( - capacity = dict() - ) - ) - - for key, value in module.params.iteritems(): - if key in ['cpu', 'memory']: - node_def['resources']['capacity'][key] = value - elif key == 'name': - node_def['id'] = value - elif key != 'client_config': - if value: - node_def[key] = value + # TODO: provide sane defaults for some (like hostname, externalIP, + # internalIP, etc) + node = Node(module, client_opts, module.params['api_version'], + module.params['name'], module.params['host_ip'], + module.params['hostnames'], module.params['external_ips'], + module.params['internal_ips'], module.params['cpu'], + module.params['memory'], module.params['labels'], + module.params['annotations'], module.params['pod_cidr'], + module.params['external_id']) - if not node_def['resources']['capacity']['cpu']: - node_def['resources']['capacity']['cpu'] = multiprocessing.cpu_count() - - if not node_def['resources']['capacity']['memory']: - with open('/proc/meminfo', 'r') as mem: - for line in mem: - entries = line.split() - if str(entries.pop(0)) == 'MemTotal:': - mem_total_kb = int(entries.pop(0)) - mem_capacity = int(mem_total_kb * 1024 * .75) - node_def['resources']['capacity']['memory'] = mem_capacity - break - - try: - output = check_output(["/usr/bin/osc", "get", "nodes"] + client_opts, - stderr=subprocess.STDOUT) - except subprocess.CalledProcessError as e: - module.fail_json(msg="Failed to get node list", command=e.cmd, - returncode=e.returncode, output=e.output) - - if re.search(module.params['name'], output, re.MULTILINE): - module.exit_json(changed=False, node_def=node_def) + # TODO: attempt to support changing node settings where possible and/or + # modifying node resources + if node.exists(): + module.exit_json(changed=False, node=node.get_node()) elif module.check_mode: - module.exit_json(changed=True, node_def=node_def) - - config_def = dict( - metadata = dict( - name = "add-node-%s" % module.params['name'] - ), - kind = 'Config', - apiVersion = 'v1beta1', - items = [node_def] - ) - - p = Popen(["/usr/bin/osc"] + client_opts + ["create", "node"] + ["-f", "-"], - stdin=subprocess.PIPE, stdout=subprocess.PIPE, - stderr=subprocess.PIPE, close_fds=True) - (out, err) = p.communicate(module.jsonify(config_def)) - ret = p.returncode - - if ret != 0: - if re.search("minion \"%s\" already exists" % module.params['name'], - err): - module.exit_json(changed=False, - msg="node definition already exists", config_def=config_def) + module.exit_json(changed=True, node=node.get_node()) + else: + if node.create(): + module.exit_json(changed=True, + msg="Node created successfully", + node=node.get_node()) else: - module.fail_json(msg="Node creation failed.", ret=ret, out=out, - err=err, config_def=config_def) + module.fail_json(msg="Unknown error creating node", + node=node.get_node()) - module.exit_json(changed=True, out=out, err=err, ret=ret, - node_def=config_def) # import module snippets from ansible.module_utils.basic import * -main() +if __name__ == '__main__': + main() diff --git a/roles/openshift_node/tasks/main.yml b/roles/openshift_node/tasks/main.yml index 6721c7401..e380ba1fb 100644 --- a/roles/openshift_node/tasks/main.yml +++ b/roles/openshift_node/tasks/main.yml @@ -21,7 +21,7 @@ lineinfile: dest: /etc/sysconfig/openshift-node regexp: '^OPTIONS=' - line: "OPTIONS=\"--master=https://{{ openshift_master_ips[0] }}:8443 --loglevel={{ openshift_node_debug_level }}\"" + line: "OPTIONS=\"--master=https://{{ openshift_master_ips[0] }}:8443 --hostname={{ openshift_hostname }} --loglevel={{ openshift_node_debug_level }}\"" notify: - restart openshift-node @@ -75,4 +75,14 @@ - name: Register node (if not already registered) openshift_register_node: name: "{{ openshift_hostname }}" - resources: "{{ openshift_node_resources }}" + api_version: v1beta1 + cpu: "{{ openshift_node_resources.cpu }}" + memory: "{{ openshift_node_resources.memory }}" + pod_cidr: "{{ openshift_node_resources.cidr }}" + host_ip: "{{ ansible_default_ipv4.address }}" + labels: "{{ openshift_node_labels }}" + annotations: "{{ openshift_node_annotations }}" + # TODO: support customizing other attributes such as: client_config, + # client_cluster, client_context, client_user + # TODO: updated for v1beta3 changes after rebase: hostnames, external_ips, + # internal_ips, external_id -- cgit v1.2.3