diff options
Diffstat (limited to 'playbooks/common/openshift-cluster/upgrades')
14 files changed, 103 insertions, 258 deletions
diff --git a/playbooks/common/openshift-cluster/upgrades/docker/restart.yml b/playbooks/common/openshift-cluster/upgrades/docker/restart.yml index d800b289b..1b418920f 100644 --- a/playbooks/common/openshift-cluster/upgrades/docker/restart.yml +++ b/playbooks/common/openshift-cluster/upgrades/docker/restart.yml @@ -19,11 +19,9 @@ when: openshift.common.is_containerized | bool - name: Wait for master API to come back online - become: no - local_action: - module: wait_for - host="{{ inventory_hostname }}" - state=started - delay=10 - port="{{ openshift.master.api_port }}" + wait_for: + host: "{{ openshift.common.hostname }}" + state: started + delay: 10 + port: "{{ openshift.master.api_port }}" when: inventory_hostname in groups.oo_masters_to_config diff --git a/playbooks/common/openshift-cluster/upgrades/docker/upgrade.yml b/playbooks/common/openshift-cluster/upgrades/docker/upgrade.yml index 44ddf97ad..17f8fc6e9 100644 --- a/playbooks/common/openshift-cluster/upgrades/docker/upgrade.yml +++ b/playbooks/common/openshift-cluster/upgrades/docker/upgrade.yml @@ -20,7 +20,7 @@ - debug: var=docker_image_count.stdout - name: Remove all containers and images - script: nuke_images.sh docker + script: nuke_images.sh register: nuke_images_result when: docker_upgrade_nuke_images is defined and docker_upgrade_nuke_images | bool diff --git a/playbooks/common/openshift-cluster/upgrades/docker/upgrade_check.yml b/playbooks/common/openshift-cluster/upgrades/docker/upgrade_check.yml index e3379f29b..b2a2eac9a 100644 --- a/playbooks/common/openshift-cluster/upgrades/docker/upgrade_check.yml +++ b/playbooks/common/openshift-cluster/upgrades/docker/upgrade_check.yml @@ -30,9 +30,9 @@ changed_when: false - fail: - msg: This playbook requires access to Docker 1.10 or later - # Disable the 1.10 requirement if the user set a specific Docker version - when: docker_version is not defined and (docker_upgrade is not defined or docker_upgrade | bool == True) and (pkg_check.rc == 0 and (avail_docker_version.stdout == "" or avail_docker_version.stdout | version_compare('1.10','<'))) + msg: This playbook requires access to Docker 1.12 or later + # Disable the 1.12 requirement if the user set a specific Docker version + when: docker_version is not defined and (docker_upgrade is not defined or docker_upgrade | bool == True) and (pkg_check.rc == 0 and (avail_docker_version.stdout == "" or avail_docker_version.stdout | version_compare('1.12','<'))) # Default l_docker_upgrade to False, we'll set to True if an upgrade is required: - set_fact: diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/backup.yml b/playbooks/common/openshift-cluster/upgrades/etcd/backup.yml index 0a972adf6..d0eadf1fc 100644 --- a/playbooks/common/openshift-cluster/upgrades/etcd/backup.yml +++ b/playbooks/common/openshift-cluster/upgrades/etcd/backup.yml @@ -4,6 +4,7 @@ vars: embedded_etcd: "{{ groups.oo_etcd_to_config | default([]) | length == 0 }}" timestamp: "{{ lookup('pipe', 'date +%Y%m%d%H%M%S') }}" + etcdctl_command: "{{ 'etcdctl' if not openshift.common.is_containerized or embedded_etcd else 'docker exec etcd_container etcdctl' }}" roles: - openshift_facts tasks: @@ -42,19 +43,32 @@ {{ avail_disk.stdout }} Kb available. when: (embedded_etcd | bool) and (etcd_disk_usage.stdout|int > avail_disk.stdout|int) - # TODO - Refactor containerized backup to use etcd_container to backup the data so we don't rely on - # the host's etcdctl binary which may be of a different version. - - # for non containerized and non embedded we should have the correct version of etcd installed already - # For embedded we need to use the latest because OCP 3.3 uses a version of etcd that can only be backed - # up with etcd-3.x + # For non containerized and non embedded we should have the correct version of + # etcd installed already. So don't do anything. + # + # For embedded or containerized we need to use the latest because OCP 3.3 uses + # a version of etcd that can only be backed up with etcd-3.x and if it's + # containerized then etcd version may be newer than that on the host so + # upgrade it. + # + # On atomic we have neither yum nor dnf so ansible throws a hard to debug error + # if you use package there, like this: "Could not find a module for unknown." + # see https://bugzilla.redhat.com/show_bug.cgi?id=1408668 + # + # TODO - We should refactor all containerized backups to use the containerized + # version of etcd to perform the backup rather than relying on the host's + # binaries. Until we do that we'll continue to have problems backing up etcd + # when atomic host has an older version than the version that's running in the + # container whether that's embedded or not - name: Install latest etcd for containerized or embedded - package: name=etcd state=latest - when: ( openshift.common.is_containerized and not openshift.common.is_atomic ) or embedded_etcd | bool + package: + name: etcd + state: latest + when: ( embedded_etcd | bool or openshift.common.is_containerized ) and not openshift.common.is_atomic - name: Generate etcd backup command: > - etcdctl backup --data-dir={{ openshift.etcd.etcd_data_dir }} + {{ etcdctl_command }} backup --data-dir={{ openshift.etcd.etcd_data_dir }} --backup-dir={{ openshift.common.data_dir }}/etcd-backup-{{ backup_tag | default('') }}{{ timestamp }} - set_fact: diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/containerized_tasks.yml b/playbooks/common/openshift-cluster/upgrades/etcd/containerized_tasks.yml index f88981a0b..5f8b59e17 100644 --- a/playbooks/common/openshift-cluster/upgrades/etcd/containerized_tasks.yml +++ b/playbooks/common/openshift-cluster/upgrades/etcd/containerized_tasks.yml @@ -8,8 +8,7 @@ - name: Set new_etcd_image set_fact: - new_etcd_image: "{{ current_image.stdout | regex_replace('/etcd.*$','/etcd3:' ~ upgrade_version ) if upgrade_version | version_compare('3.0','>=') - else current_image.stdout.split(':')[0] ~ ':' ~ upgrade_version }}" + new_etcd_image: "{{ current_image.stdout | regex_replace('/etcd.*$','/etcd:' ~ upgrade_version ) }}" - name: Pull new etcd image command: "docker pull {{ new_etcd_image }}" diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/upgrade.yml b/playbooks/common/openshift-cluster/upgrades/etcd/upgrade.yml index 5ff9521ec..0f8d94737 100644 --- a/playbooks/common/openshift-cluster/upgrades/etcd/upgrade.yml +++ b/playbooks/common/openshift-cluster/upgrades/etcd/upgrade.yml @@ -75,7 +75,7 @@ hosts: etcd_hosts_to_upgrade serial: 1 vars: - upgrade_version: 3.0.14 + upgrade_version: 3.0.15 tasks: - include: containerized_tasks.yml when: etcd_container_version.stdout | default('99') | version_compare('3.0','<') and openshift.common.is_containerized | bool diff --git a/playbooks/common/openshift-cluster/upgrades/files/pre-upgrade-check b/playbooks/common/openshift-cluster/upgrades/files/pre-upgrade-check deleted file mode 100644 index e5c958ebb..000000000 --- a/playbooks/common/openshift-cluster/upgrades/files/pre-upgrade-check +++ /dev/null @@ -1,193 +0,0 @@ -#!/usr/bin/env python -""" -Pre-upgrade checks that must be run on a master before proceeding with upgrade. -""" -# This is a script not a python module: -# pylint: disable=invalid-name - -# NOTE: This script should not require any python libs other than what is -# in the standard library. - -__license__ = "ASL 2.0" - -import json -import os -import subprocess -import re - -# The maximum length of container.ports.name -ALLOWED_LENGTH = 15 -# The valid structure of container.ports.name -ALLOWED_CHARS = re.compile('^[a-z0-9][a-z0-9\\-]*[a-z0-9]$') -AT_LEAST_ONE_LETTER = re.compile('[a-z]') -# look at OS_PATH for the full path. Default ot 'oc' -OC_PATH = os.getenv('OC_PATH', 'oc') - - -def validate(value): - """ - validate verifies that value matches required conventions - - Rules of container.ports.name validation: - - * must be less that 16 chars - * at least one letter - * only a-z0-9- - * hyphens can not be leading or trailing or next to each other - - :Parameters: - - `value`: Value to validate - """ - if len(value) > ALLOWED_LENGTH: - return False - - if '--' in value: - return False - - # We search since it can be anywhere - if not AT_LEAST_ONE_LETTER.search(value): - return False - - # We match because it must start at the beginning - if not ALLOWED_CHARS.match(value): - return False - return True - - -def list_items(kind): - """ - list_items returns a list of items from the api - - :Parameters: - - `kind`: Kind of item to access - """ - response = subprocess.check_output([OC_PATH, 'get', '--all-namespaces', '-o', 'json', kind]) - items = json.loads(response) - return items.get("items", []) - - -def get(obj, *paths): - """ - Gets an object - - :Parameters: - - `obj`: A dictionary structure - - `path`: All other non-keyword arguments - """ - ret_obj = obj - for path in paths: - if ret_obj.get(path, None) is None: - return [] - ret_obj = ret_obj[path] - return ret_obj - - -# pylint: disable=too-many-arguments -def pretty_print_errors(namespace, kind, item_name, container_name, invalid_label, port_name, valid): - """ - Prints out results in human friendly way. - - :Parameters: - - `namespace`: Namespace of the resource - - `kind`: Kind of the resource - - `item_name`: Name of the resource - - `container_name`: Name of the container. May be "" when kind=Service. - - `port_name`: Name of the port - - `invalid_label`: The label of the invalid port. Port.name/targetPort - - `valid`: True if the port is valid - """ - if not valid: - if len(container_name) > 0: - print('%s/%s -n %s (Container="%s" %s="%s")' % ( - kind, item_name, namespace, container_name, invalid_label, port_name)) - else: - print('%s/%s -n %s (%s="%s")' % ( - kind, item_name, namespace, invalid_label, port_name)) - - -def print_validation_header(): - """ - Prints the error header. Should run on the first error to avoid - overwhelming the user. - """ - print """\ -At least one port name is invalid and must be corrected before upgrading. -Please update or remove any resources with invalid port names. - - Valid port names must: - - * be less that 16 characters - * have at least one letter - * contain only a-z0-9- - * not start or end with - - * not contain dashes next to each other ('--') -""" - - -def main(): - """ - main is the main entry point to this script - """ - try: - # the comma at the end suppresses the newline - print "Checking for oc ...", - subprocess.check_output([OC_PATH, 'whoami']) - print "found" - except: - print( - 'Unable to run "%s whoami"\n' - 'Please ensure OpenShift is running, and "oc" is on your system ' - 'path.\n' - 'You can override the path with the OC_PATH environment variable.' - % OC_PATH) - raise SystemExit(1) - - # Where the magic happens - first_error = True - for kind, path in [ - ('deploymentconfigs', ("spec", "template", "spec", "containers")), - ('replicationcontrollers', ("spec", "template", "spec", "containers")), - ('pods', ("spec", "containers"))]: - for item in list_items(kind): - namespace = item["metadata"]["namespace"] - item_name = item["metadata"]["name"] - for container in get(item, *path): - container_name = container["name"] - for port in get(container, "ports"): - port_name = port.get("name", None) - if not port_name: - # Unnamed ports are OK - continue - valid = validate(port_name) - if not valid and first_error: - first_error = False - print_validation_header() - pretty_print_errors( - namespace, kind, item_name, - container_name, "Port.name", port_name, valid) - - # Services follow a different flow - for item in list_items('services'): - namespace = item["metadata"]["namespace"] - item_name = item["metadata"]["name"] - for port in get(item, "spec", "ports"): - port_name = port.get("targetPort", None) - if isinstance(port_name, int) or port_name is None: - # Integer only or unnamed ports are OK - continue - valid = validate(port_name) - if not valid and first_error: - first_error = False - print_validation_header() - pretty_print_errors( - namespace, "services", item_name, "", - "targetPort", port_name, valid) - - # If we had at least 1 error then exit with 1 - if not first_error: - raise SystemExit(1) - - -if __name__ == '__main__': - main() - diff --git a/playbooks/common/openshift-cluster/upgrades/files/rpm_versions.sh b/playbooks/common/openshift-cluster/upgrades/files/rpm_versions.sh deleted file mode 100644 index 7bf249742..000000000 --- a/playbooks/common/openshift-cluster/upgrades/files/rpm_versions.sh +++ /dev/null @@ -1,12 +0,0 @@ -#!/bin/bash -if [ `which dnf 2> /dev/null` ]; then - installed=$(dnf repoquery --installed --latest-limit 1 -d 0 --qf '%{version}-%{release}' "${@}" 2> /dev/null) - available=$(dnf repoquery --available --latest-limit 1 -d 0 --qf '%{version}-%{release}' "${@}" 2> /dev/null) -else - installed=$(repoquery --plugins --pkgnarrow=installed --qf '%{version}-%{release}' "${@}" 2> /dev/null) - available=$(repoquery --plugins --pkgnarrow=available --qf '%{version}-%{release}' "${@}" 2> /dev/null) -fi - -echo "---" -echo "curr_version: ${installed}" -echo "avail_version: ${available}" diff --git a/playbooks/common/openshift-cluster/upgrades/init.yml b/playbooks/common/openshift-cluster/upgrades/init.yml index 8cac2fb3b..76645ff3f 100644 --- a/playbooks/common/openshift-cluster/upgrades/init.yml +++ b/playbooks/common/openshift-cluster/upgrades/init.yml @@ -1,5 +1,6 @@ --- -- hosts: localhost +- name: Create l_oo_all_hosts group + hosts: localhost connection: local become: no gather_facts: no @@ -10,7 +11,8 @@ groups: l_oo_all_hosts with_items: "{{ g_all_hosts | default([]) }}" -- hosts: l_oo_all_hosts +- name: Include g_*_hosts vars for hosts in group l_oo_all_hosts + hosts: l_oo_all_hosts gather_facts: no tasks: - include_vars: ../../../byo/openshift-cluster/cluster_hosts.yml @@ -46,3 +48,14 @@ when: openshift_docker_log_options is not defined - include: ../initialize_facts.yml + +- name: Ensure clean repo cache in the event repos have been changed manually + hosts: oo_all_hosts + tags: + - pre_upgrade + tasks: + - name: Clean package cache + command: "{{ ansible_pkg_mgr }} clean all" + when: not openshift.common.is_atomic | bool + args: + warn: no diff --git a/playbooks/common/openshift-cluster/upgrades/library/openshift_upgrade_config.py b/playbooks/common/openshift-cluster/upgrades/library/openshift_upgrade_config.py index 1238acb05..673f11889 100755 --- a/playbooks/common/openshift-cluster/upgrades/library/openshift_upgrade_config.py +++ b/playbooks/common/openshift-cluster/upgrades/library/openshift_upgrade_config.py @@ -146,7 +146,7 @@ def main(): # ignore broad-except error to avoid stack trace to ansible user # pylint: disable=broad-except - except Exception, e: + except Exception as e: return module.fail_json(msg=str(e)) diff --git a/playbooks/common/openshift-cluster/upgrades/pre/verify_docker_upgrade_targets.yml b/playbooks/common/openshift-cluster/upgrades/pre/verify_docker_upgrade_targets.yml index ba4d77617..7646e0fa6 100644 --- a/playbooks/common/openshift-cluster/upgrades/pre/verify_docker_upgrade_targets.yml +++ b/playbooks/common/openshift-cluster/upgrades/pre/verify_docker_upgrade_targets.yml @@ -19,5 +19,5 @@ when: openshift.common.is_atomic | bool - fail: - msg: This playbook requires access to Docker 1.10 or later - when: openshift.common.is_atomic | bool and l_docker_version.avail_version | default(l_docker_version.curr_version, true) | version_compare('1.10','<') + msg: This playbook requires access to Docker 1.12 or later + when: openshift.common.is_atomic | bool and l_docker_version.avail_version | default(l_docker_version.curr_version, true) | version_compare('1.12','<') diff --git a/playbooks/common/openshift-cluster/upgrades/pre/verify_upgrade_targets.yml b/playbooks/common/openshift-cluster/upgrades/pre/verify_upgrade_targets.yml index 9632626a4..c83923dae 100644 --- a/playbooks/common/openshift-cluster/upgrades/pre/verify_upgrade_targets.yml +++ b/playbooks/common/openshift-cluster/upgrades/pre/verify_upgrade_targets.yml @@ -12,10 +12,6 @@ msg: Verify the correct version was found when: verify_upgrade_version is defined and openshift_version != verify_upgrade_version - - name: Clean package cache - command: "{{ ansible_pkg_mgr }} clean all" - when: not openshift.common.is_atomic | bool - - set_fact: g_new_service_name: "{{ 'origin' if deployment_type =='origin' else 'atomic-openshift' }}" when: not openshift.common.is_containerized | bool diff --git a/playbooks/common/openshift-cluster/upgrades/upgrade_control_plane.yml b/playbooks/common/openshift-cluster/upgrades/upgrade_control_plane.yml index 474e6311e..77b37cdc2 100644 --- a/playbooks/common/openshift-cluster/upgrades/upgrade_control_plane.yml +++ b/playbooks/common/openshift-cluster/upgrades/upgrade_control_plane.yml @@ -30,14 +30,6 @@ - name: Upgrade and backup etcd include: ./etcd/main.yml -- name: Upgrade master packages - hosts: oo_masters_to_config - roles: - - openshift_facts - tasks: - - include: rpm_upgrade.yml component=master - when: not openshift.common.is_containerized | bool - # Create service signer cert when missing. Service signer certificate # is added to master config in the master config hook for v3_3. - name: Determine if service signer cert must be created @@ -51,15 +43,40 @@ - include: create_service_signer_cert.yml -- name: Upgrade master config and systemd units +# Set openshift_master_facts separately. In order to reconcile +# admission_config's, we currently must run openshift_master_facts and +# then run openshift_facts. +- name: Set OpenShift master facts + hosts: oo_masters_to_config + roles: + - openshift_master_facts + +# The main master upgrade play. Should handle all changes to the system in one pass, with +# support for optional hooks to be defined. +- name: Upgrade master hosts: oo_masters_to_config + vars: + openshift_master_ha: "{{ groups.oo_masters_to_config | length > 1 }}" + serial: 1 handlers: - include: ../../../../roles/openshift_master/handlers/main.yml static: yes roles: - openshift_facts - - openshift_master_facts - tasks: + post_tasks: + + # Run the pre-upgrade hook if defined: + - debug: msg="Running master pre-upgrade hook {{ openshift_master_upgrade_pre_hook }}" + when: openshift_master_upgrade_pre_hook is defined + + - include: "{{ openshift_master_upgrade_pre_hook }}" + when: openshift_master_upgrade_pre_hook is defined + + - include: rpm_upgrade.yml component=master + when: not openshift.common.is_containerized | bool + + - include_vars: ../../../../roles/openshift_master_facts/vars/main.yml + - include: upgrade_scheduler.yml - include: "{{ master_config_hook }}" @@ -95,9 +112,26 @@ state: link when: ca_crt_stat.stat.isreg and not ca_bundle_stat.stat.exists -- name: Set master update status to complete - hosts: oo_masters_to_config - tasks: + # Run the upgrade hook prior to restarting services/system if defined: + - debug: msg="Running master upgrade hook {{ openshift_master_upgrade_hook }}" + when: openshift_master_upgrade_hook is defined + + - include: "{{ openshift_master_upgrade_hook }}" + when: openshift_master_upgrade_hook is defined + + - include: ../../openshift-master/restart_hosts.yml + when: openshift.common.rolling_restart_mode == 'system' + + - include: ../../openshift-master/restart_services.yml + when: openshift.common.rolling_restart_mode == 'services' + + # Run the post-upgrade hook if defined: + - debug: msg="Running master post-upgrade hook {{ openshift_master_upgrade_post_hook }}" + when: openshift_master_upgrade_post_hook is defined + + - include: "{{ openshift_master_upgrade_post_hook }}" + when: openshift_master_upgrade_post_hook is defined + - set_fact: master_update_complete: True @@ -119,10 +153,6 @@ msg: "Upgrade cannot continue. The following masters did not finish updating: {{ master_update_failed | join(',') }}" when: master_update_failed | length > 0 -# We are now ready to restart master services (or entire system -# depending on openshift_rolling_restart_mode): -- include: ../../openshift-master/restart.yml - ############################################################################### # Reconcile Cluster Roles, Cluster Role Bindings and Security Context Constraints ############################################################################### diff --git a/playbooks/common/openshift-cluster/upgrades/upgrade_nodes.yml b/playbooks/common/openshift-cluster/upgrades/upgrade_nodes.yml index b3ac34d90..2bb460815 100644 --- a/playbooks/common/openshift-cluster/upgrades/upgrade_nodes.yml +++ b/playbooks/common/openshift-cluster/upgrades/upgrade_nodes.yml @@ -1,5 +1,5 @@ --- -- name: Evacuate and upgrade nodes +- name: Drain and upgrade nodes hosts: oo_nodes_to_upgrade # This var must be set with -e on invocation, as it is not a per-host inventory var # and is evaluated early. Values such as "20%" can also be used. @@ -39,9 +39,9 @@ retries: 3 delay: 1 - - name: Evacuate Node for Kubelet upgrade + - name: Drain Node for Kubelet upgrade command: > - {{ hostvars[groups.oo_first_master.0].openshift.common.client_binary }} adm manage-node {{ openshift.node.nodename | lower }} --evacuate --force + {{ hostvars[groups.oo_first_master.0].openshift.common.client_binary }} adm manage-node {{ openshift.node.nodename | lower }} {{ openshift.common.evacuate_or_drain }} --force delegate_to: "{{ groups.oo_first_master.0 }}" when: inventory_hostname in groups.oo_nodes_to_upgrade |