diff options
Diffstat (limited to 'playbooks/common/openshift-cluster/upgrades/etcd')
9 files changed, 137 insertions, 298 deletions
diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/backup.yml b/playbooks/common/openshift-cluster/upgrades/etcd/backup.yml index 7ef79afa9..616ba04f8 100644 --- a/playbooks/common/openshift-cluster/upgrades/etcd/backup.yml +++ b/playbooks/common/openshift-cluster/upgrades/etcd/backup.yml @@ -1,88 +1,14 @@ --- - name: Backup etcd - hosts: etcd_hosts_to_backup - 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' if not openshift.common.is_etcd_system_container else 'runc exec etcd etcdctl' }}" + hosts: oo_etcd_hosts_to_backup roles: - - openshift_facts - tasks: - # Ensure we persist the etcd role for this host in openshift_facts - - openshift_facts: - role: etcd - local_facts: {} - when: "'etcd' not in openshift" - - - stat: path=/var/lib/openshift - register: var_lib_openshift - - - stat: path=/var/lib/origin - register: var_lib_origin - - - name: Create origin symlink if necessary - file: src=/var/lib/openshift/ dest=/var/lib/origin state=link - when: var_lib_openshift.stat.exists == True and var_lib_origin.stat.exists == False - - # TODO: replace shell module with command and update later checks - # We assume to be using the data dir for all backups. - - name: Check available disk space for etcd backup - shell: df --output=avail -k {{ openshift.common.data_dir }} | tail -n 1 - register: avail_disk - # AUDIT:changed_when: `false` because we are only inspecting - # state, not manipulating anything - changed_when: false - - # TODO: replace shell module with command and update later checks - - name: Check current embedded etcd disk usage - shell: du -k {{ openshift.etcd.etcd_data_dir }} | tail -n 1 | cut -f1 - register: etcd_disk_usage - when: embedded_etcd | bool - # AUDIT:changed_when: `false` because we are only inspecting - # state, not manipulating anything - changed_when: false - - - name: Abort if insufficient disk space for etcd backup - fail: - msg: > - {{ etcd_disk_usage.stdout }} Kb disk space required for etcd backup, - {{ avail_disk.stdout }} Kb available. - when: (embedded_etcd | bool) and (etcd_disk_usage.stdout|int > avail_disk.stdout|int) - - # 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: ( embedded_etcd | bool or openshift.common.is_containerized ) and not openshift.common.is_atomic - - - name: Generate etcd backup - command: > - {{ etcdctl_command }} backup --data-dir={{ openshift.etcd.etcd_data_dir }} - --backup-dir={{ openshift.common.data_dir }}/etcd-backup-{{ backup_tag | default('') }}{{ timestamp }} - - - set_fact: - etcd_backup_complete: True - - - name: Display location of etcd backup - debug: - msg: "Etcd backup created in {{ openshift.common.data_dir }}/etcd-backup-{{ backup_tag | default('') }}{{ timestamp }}" + - role: openshift_facts + - role: etcd_common + r_etcd_common_action: backup + r_etcd_common_backup_tag: etcd_backup_tag + r_etcd_common_etcd_runtime: "{{ openshift.common.etcd_runtime }}" + r_etcd_common_embedded_etcd: "{{ groups.oo_etcd_to_config | default([]) | length == 0 }}" + r_etcd_common_backup_sufix_name: "{{ lookup('pipe', 'date +%Y%m%d%H%M%S') }}" - name: Gate on etcd backup hosts: localhost @@ -91,10 +17,10 @@ tasks: - set_fact: etcd_backup_completed: "{{ hostvars - | oo_select_keys(groups.etcd_hosts_to_backup) - | oo_collect('inventory_hostname', {'etcd_backup_complete': true}) }}" + | oo_select_keys(groups.oo_etcd_hosts_to_backup) + | oo_collect('inventory_hostname', {'r_etcd_common_backup_complete': true}) }}" - set_fact: - etcd_backup_failed: "{{ groups.etcd_hosts_to_backup | difference(etcd_backup_completed) }}" + etcd_backup_failed: "{{ groups.oo_etcd_hosts_to_backup | difference(etcd_backup_completed) }}" - fail: msg: "Upgrade cannot continue. The following hosts did not complete etcd backup: {{ etcd_backup_failed | join(',') }}" when: etcd_backup_failed | length > 0 diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/containerized_tasks.yml b/playbooks/common/openshift-cluster/upgrades/etcd/containerized_tasks.yml deleted file mode 100644 index 5f8b59e17..000000000 --- a/playbooks/common/openshift-cluster/upgrades/etcd/containerized_tasks.yml +++ /dev/null @@ -1,46 +0,0 @@ ---- -- name: Verify cluster is healthy pre-upgrade - command: "etcdctl --cert-file /etc/etcd/peer.crt --key-file /etc/etcd/peer.key --ca-file /etc/etcd/ca.crt -C https://{{ openshift.common.hostname }}:2379 cluster-health" - -- name: Get current image - shell: grep 'ExecStart=' /etc/systemd/system/etcd_container.service | awk '{print $NF}' - register: current_image - -- name: Set new_etcd_image - set_fact: - new_etcd_image: "{{ current_image.stdout | regex_replace('/etcd.*$','/etcd:' ~ upgrade_version ) }}" - -- name: Pull new etcd image - command: "docker pull {{ new_etcd_image }}" - -- name: Update to latest etcd image - replace: - dest: /etc/systemd/system/etcd_container.service - regexp: "{{ current_image.stdout }}$" - replace: "{{ new_etcd_image }}" - -- name: Restart etcd_container - systemd: - name: etcd_container - daemon_reload: yes - state: restarted - -## TODO: probably should just move this into the backup playbooks, also this -## will fail on atomic host. We need to revisit how to do etcd backups there as -## the container may be newer than etcdctl on the host. Assumes etcd3 obsoletes etcd (7.3.1) -- name: Upgrade etcd for etcdctl when not atomic - package: name=etcd state=latest - when: not openshift.common.is_atomic | bool - -- name: Verify cluster is healthy - command: "etcdctl --cert-file /etc/etcd/peer.crt --key-file /etc/etcd/peer.key --ca-file /etc/etcd/ca.crt -C https://{{ openshift.common.hostname }}:2379 cluster-health" - register: etcdctl - until: etcdctl.rc == 0 - retries: 3 - delay: 10 - -- name: Store new etcd_image - openshift_facts: - role: etcd - local_facts: - etcd_image: "{{ new_etcd_image }}" diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/fedora_tasks.yml b/playbooks/common/openshift-cluster/upgrades/etcd/fedora_tasks.yml deleted file mode 100644 index 30232110e..000000000 --- a/playbooks/common/openshift-cluster/upgrades/etcd/fedora_tasks.yml +++ /dev/null @@ -1,23 +0,0 @@ ---- -# F23 GA'd with etcd 2.0, currently has 2.2 in updates -# F24 GA'd with etcd-2.2, currently has 2.2 in updates -# F25 Beta currently has etcd 3.0 -- name: Verify cluster is healthy pre-upgrade - command: "etcdctl --cert-file /etc/etcd/peer.crt --key-file /etc/etcd/peer.key --ca-file /etc/etcd/ca.crt -C https://{{ openshift.common.hostname }}:2379 cluster-health" - -- name: Update etcd - package: - name: "etcd" - state: "latest" - -- name: Restart etcd - service: - name: etcd - state: restarted - -- name: Verify cluster is healthy - command: "etcdctl --cert-file /etc/etcd/peer.crt --key-file /etc/etcd/peer.key --ca-file /etc/etcd/ca.crt -C https://{{ openshift.common.hostname }}:2379 cluster-health" - register: etcdctl - until: etcdctl.rc == 0 - retries: 3 - delay: 10 diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/files/etcdctl.sh b/playbooks/common/openshift-cluster/upgrades/etcd/files/etcdctl.sh deleted file mode 120000 index 641e04e44..000000000 --- a/playbooks/common/openshift-cluster/upgrades/etcd/files/etcdctl.sh +++ /dev/null @@ -1 +0,0 @@ -../roles/etcd/files/etcdctl.sh
\ No newline at end of file diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/main.yml b/playbooks/common/openshift-cluster/upgrades/etcd/main.yml index fa86d29fb..64abc54e7 100644 --- a/playbooks/common/openshift-cluster/upgrades/etcd/main.yml +++ b/playbooks/common/openshift-cluster/upgrades/etcd/main.yml @@ -5,42 +5,19 @@ # mirrored packages on your own because only the GA and latest versions are # available in the repos. So for Fedora we'll simply skip this, sorry. -- include: ../../evaluate_groups.yml - tags: - - always - -# We use two groups one for hosts we're upgrading which doesn't include embedded etcd -# The other for backing up which includes the embedded etcd host, there's no need to -# upgrade embedded etcd that just happens when the master is updated. -- name: Evaluate additional groups for etcd - hosts: localhost - connection: local - become: no - tasks: - - name: Evaluate etcd_hosts_to_upgrade - add_host: - name: "{{ item }}" - groups: etcd_hosts_to_upgrade - with_items: "{{ groups.oo_etcd_to_config if groups.oo_etcd_to_config is defined and groups.oo_etcd_to_config | length > 0 else [] }}" - changed_when: False - - - name: Evaluate etcd_hosts_to_backup - add_host: - name: "{{ item }}" - groups: etcd_hosts_to_backup - with_items: "{{ groups.oo_etcd_to_config if groups.oo_etcd_to_config is defined and groups.oo_etcd_to_config | length > 0 else groups.oo_first_master }}" - changed_when: False - - name: Backup etcd before upgrading anything include: backup.yml vars: - backup_tag: "pre-upgrade-" + etcd_backup_tag: "pre-upgrade-" when: openshift_etcd_backup | default(true) | bool - name: Drop etcdctl profiles - hosts: etcd_hosts_to_upgrade + hosts: oo_etcd_hosts_to_upgrade tasks: - - include: roles/etcd/tasks/etcdctl.yml + - include_role: + name: etcd_common + vars: + r_etcd_common_action: drop_etcdctl - name: Perform etcd upgrade include: ./upgrade.yml diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/rhel_tasks.yml b/playbooks/common/openshift-cluster/upgrades/etcd/rhel_tasks.yml deleted file mode 100644 index 3a972e8ab..000000000 --- a/playbooks/common/openshift-cluster/upgrades/etcd/rhel_tasks.yml +++ /dev/null @@ -1,20 +0,0 @@ ---- -- name: Verify cluster is healthy pre-upgrade - command: "etcdctl --cert-file /etc/etcd/peer.crt --key-file /etc/etcd/peer.key --ca-file /etc/etcd/ca.crt -C https://{{ openshift.common.hostname }}:2379 cluster-health" - -- name: Update etcd RPM - package: - name: etcd-{{ upgrade_version }}* - state: latest - -- name: Restart etcd - service: - name: etcd - state: restarted - -- name: Verify cluster is healthy - command: "etcdctl --cert-file /etc/etcd/peer.crt --key-file /etc/etcd/peer.key --ca-file /etc/etcd/ca.crt -C https://{{ openshift.common.hostname }}:2379 cluster-health" - register: etcdctl - until: etcdctl.rc == 0 - retries: 3 - delay: 10 diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/upgrade.yml b/playbooks/common/openshift-cluster/upgrades/etcd/upgrade.yml index a9b5b94e6..39e82498d 100644 --- a/playbooks/common/openshift-cluster/upgrades/etcd/upgrade.yml +++ b/playbooks/common/openshift-cluster/upgrades/etcd/upgrade.yml @@ -1,119 +1,110 @@ --- - name: Determine etcd version - hosts: etcd_hosts_to_upgrade + hosts: oo_etcd_hosts_to_upgrade tasks: - - name: Record RPM based etcd version - command: rpm -qa --qf '%{version}' etcd\* - args: - warn: no - register: etcd_rpm_version - failed_when: false - when: not openshift.common.is_containerized | bool - # AUDIT:changed_when: `false` because we are only inspecting - # state, not manipulating anything - changed_when: false - - - name: Record containerized etcd version - command: docker exec etcd_container rpm -qa --qf '%{version}' etcd\* - register: etcd_container_version - failed_when: false - when: openshift.common.is_containerized | bool - # AUDIT:changed_when: `false` because we are only inspecting - # state, not manipulating anything - changed_when: false - - - name: Record containerized etcd version - command: docker exec etcd_container rpm -qa --qf '%{version}' etcd\* - register: etcd_container_version - failed_when: false - when: openshift.common.is_containerized | bool and not openshift.common.is_etcd_system_container | bool - # AUDIT:changed_when: `false` because we are only inspecting - # state, not manipulating anything - changed_when: false - - - name: Record containerized etcd version - command: runc exec etcd_container rpm -qa --qf '%{version}' etcd\* - register: etcd_container_version - failed_when: false - when: openshift.common.is_containerized | bool and openshift.common.is_etcd_system_container | bool - # AUDIT:changed_when: `false` because we are only inspecting - # state, not manipulating anything - changed_when: false - -# I really dislike this copy/pasta but I wasn't able to find a way to get it to loop -# through hosts, then loop through tasks only when appropriate -- name: Upgrade to 2.1 - hosts: etcd_hosts_to_upgrade - serial: 1 + - block: + - name: Record RPM based etcd version + command: rpm -qa --qf '%{version}' etcd\* + args: + warn: no + register: etcd_rpm_version + failed_when: false + # AUDIT:changed_when: `false` because we are only inspecting + # state, not manipulating anything + changed_when: false + - debug: + msg: "Etcd rpm version {{ etcd_rpm_version.stdout }} detected" + when: + - not openshift.common.is_containerized | bool + + - block: + - name: Record containerized etcd version (docker) + command: docker exec etcd_container rpm -qa --qf '%{version}' etcd\* + register: etcd_container_version_docker + failed_when: false + # AUDIT:changed_when: `false` because we are only inspecting + # state, not manipulating anything + changed_when: false + when: + - not openshift.common.is_etcd_system_container | bool + + # Given a register variables is set even if the whwen condition + # is false, we need to set etcd_container_version separately + - set_fact: + etcd_container_version: "{{ etcd_container_version_docker.stdout }}" + when: + - not openshift.common.is_etcd_system_container | bool + + - name: Record containerized etcd version (runc) + command: runc exec etcd rpm -qa --qf '%{version}' etcd\* + register: etcd_container_version_runc + failed_when: false + # AUDIT:changed_when: `false` because we are only inspecting + # state, not manipulating anything + changed_when: false + when: + - openshift.common.is_etcd_system_container | bool + + # Given a register variables is set even if the whwen condition + # is false, we need to set etcd_container_version separately + - set_fact: + etcd_container_version: "{{ etcd_container_version_runc.stdout }}" + when: + - openshift.common.is_etcd_system_container | bool + + - debug: + msg: "Etcd containerized version {{ etcd_container_version }} detected" + when: + - openshift.common.is_containerized | bool + +- include: upgrade_rpm_members.yml vars: - upgrade_version: '2.1' - tasks: - - include: rhel_tasks.yml - when: etcd_rpm_version.stdout | default('99') | version_compare('2.1','<') and ansible_distribution == 'RedHat' and not openshift.common.is_containerized | bool + etcd_upgrade_version: '2.1' -- name: Upgrade RPM hosts to 2.2 - hosts: etcd_hosts_to_upgrade - serial: 1 +- include: upgrade_rpm_members.yml vars: - upgrade_version: '2.2' - tasks: - - include: rhel_tasks.yml - when: etcd_rpm_version.stdout | default('99') | version_compare('2.2','<') and ansible_distribution == 'RedHat' and not openshift.common.is_containerized | bool + etcd_upgrade_version: '2.2' -- name: Upgrade containerized hosts to 2.2.5 - hosts: etcd_hosts_to_upgrade - serial: 1 +- include: upgrade_image_members.yml vars: - upgrade_version: 2.2.5 - tasks: - - include: containerized_tasks.yml - when: etcd_container_version.stdout | default('99') | version_compare('2.2','<') and openshift.common.is_containerized | bool + etcd_upgrade_version: '2.2.5' -- name: Upgrade RPM hosts to 2.3 - hosts: etcd_hosts_to_upgrade - serial: 1 +- include: upgrade_rpm_members.yml vars: - upgrade_version: '2.3' - tasks: - - include: rhel_tasks.yml - when: etcd_rpm_version.stdout | default('99') | version_compare('2.3','<') and ansible_distribution == 'RedHat' and not openshift.common.is_containerized | bool + etcd_upgrade_version: '2.3' -- name: Upgrade containerized hosts to 2.3.7 - hosts: etcd_hosts_to_upgrade - serial: 1 +- include: upgrade_image_members.yml vars: - upgrade_version: 2.3.7 - tasks: - - include: containerized_tasks.yml - when: etcd_container_version.stdout | default('99') | version_compare('2.3','<') and openshift.common.is_containerized | bool + etcd_upgrade_version: '2.3.7' -- name: Upgrade RPM hosts to 3.0 - hosts: etcd_hosts_to_upgrade - serial: 1 +- include: upgrade_rpm_members.yml vars: - upgrade_version: '3.0' - tasks: - - include: rhel_tasks.yml - when: etcd_rpm_version.stdout | default('99') | version_compare('3.0','<') and ansible_distribution == 'RedHat' and not openshift.common.is_containerized | bool + etcd_upgrade_version: '3.0' -- name: Upgrade containerized hosts to etcd3 image - hosts: etcd_hosts_to_upgrade - serial: 1 +- include: upgrade_image_members.yml vars: - 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 + etcd_upgrade_version: '3.0.15' + +- include: upgrade_rpm_members.yml + vars: + etcd_upgrade_version: '3.1' + +- include: upgrade_image_members.yml + vars: + etcd_upgrade_version: '3.1.3' - name: Upgrade fedora to latest - hosts: etcd_hosts_to_upgrade + hosts: oo_etcd_hosts_to_upgrade serial: 1 tasks: - - include: fedora_tasks.yml - when: ansible_distribution == 'Fedora' and not openshift.common.is_containerized | bool + - include_role: + name: etcd_upgrade + when: + - ansible_distribution == 'Fedora' + - not openshift.common.is_containerized | bool - name: Backup etcd include: backup.yml vars: - backup_tag: "post-3.0-" + etcd_backup_tag: "post-3.0-" when: openshift_etcd_backup | default(true) | bool diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/upgrade_image_members.yml b/playbooks/common/openshift-cluster/upgrades/etcd/upgrade_image_members.yml new file mode 100644 index 000000000..831ca8f57 --- /dev/null +++ b/playbooks/common/openshift-cluster/upgrades/etcd/upgrade_image_members.yml @@ -0,0 +1,17 @@ +--- +# INPUT etcd_upgrade_version +# INPUT etcd_container_version +# INPUT openshift.common.is_containerized +- name: Upgrade containerized hosts to {{ etcd_upgrade_version }} + hosts: oo_etcd_hosts_to_upgrade + serial: 1 + roles: + - role: etcd_upgrade + r_etcd_upgrade_action: upgrade + r_etcd_upgrade_mechanism: image + r_etcd_upgrade_version: "{{ etcd_upgrade_version }}" + r_etcd_common_etcd_runtime: "{{ openshift.common.etcd_runtime }}" + etcd_peer: "{{ openshift.common.hostname }}" + when: + - etcd_container_version | default('99') | version_compare(etcd_upgrade_version,'<') + - openshift.common.is_containerized | bool diff --git a/playbooks/common/openshift-cluster/upgrades/etcd/upgrade_rpm_members.yml b/playbooks/common/openshift-cluster/upgrades/etcd/upgrade_rpm_members.yml new file mode 100644 index 000000000..2e79451e0 --- /dev/null +++ b/playbooks/common/openshift-cluster/upgrades/etcd/upgrade_rpm_members.yml @@ -0,0 +1,18 @@ +--- +# INPUT etcd_upgrade_version +# INPUT etcd_rpm_version +# INPUT openshift.common.is_containerized +- name: Upgrade to {{ etcd_upgrade_version }} + hosts: oo_etcd_hosts_to_upgrade + serial: 1 + roles: + - role: etcd_upgrade + r_etcd_upgrade_action: upgrade + r_etcd_upgrade_mechanism: rpm + r_etcd_upgrade_version: "{{ etcd_upgrade_version }}" + r_etcd_common_etcd_runtime: "host" + etcd_peer: "{{ openshift.common.hostname }}" + when: + - etcd_rpm_version.stdout | default('99') | version_compare(etcd_upgrade_version, '<') + - ansible_distribution == 'RedHat' + - not openshift.common.is_containerized | bool |