diff options
| -rw-r--r-- | environments/hyperconverged-ceph.yaml | 3 | ||||
| -rwxr-xr-x | extraconfig/tasks/major_upgrade_controller_pacemaker_2.sh | 1 | ||||
| -rw-r--r-- | extraconfig/tasks/major_upgrade_pacemaker.yaml | 4 | ||||
| -rw-r--r-- | puppet/services/README.rst | 4 | ||||
| -rw-r--r-- | puppet/services/ceph-mon.yaml | 43 | ||||
| -rw-r--r-- | puppet/services/ceph-osd.yaml | 50 | ||||
| -rw-r--r-- | puppet/services/ceph-rgw.yaml | 4 | ||||
| -rw-r--r-- | puppet/services/nova-api.yaml | 3 | ||||
| -rw-r--r-- | puppet/services/nova-conductor.yaml | 5 | ||||
| -rw-r--r-- | puppet/services/nova-scheduler.yaml | 3 |
10 files changed, 86 insertions, 34 deletions
diff --git a/environments/hyperconverged-ceph.yaml b/environments/hyperconverged-ceph.yaml index 77fa5a49..3738072c 100644 --- a/environments/hyperconverged-ceph.yaml +++ b/environments/hyperconverged-ceph.yaml @@ -11,6 +11,7 @@ parameter_defaults: - OS::TripleO::Services::Timezone - OS::TripleO::Services::Ntp - OS::TripleO::Services::Snmp + - OS::TripleO::Services::Sshd - OS::TripleO::Services::NovaCompute - OS::TripleO::Services::NovaLibvirt - OS::TripleO::Services::Kernel @@ -25,4 +26,6 @@ parameter_defaults: - OS::TripleO::Services::OpenDaylightOvs - OS::TripleO::Services::SensuClient - OS::TripleO::Services::FluentdClient + - OS::TripleO::Services::AuditD + - OS::TripleO::Services::Collectd - OS::TripleO::Services::CephOSD diff --git a/extraconfig/tasks/major_upgrade_controller_pacemaker_2.sh b/extraconfig/tasks/major_upgrade_controller_pacemaker_2.sh index 6bfe1239..8b900842 100755 --- a/extraconfig/tasks/major_upgrade_controller_pacemaker_2.sh +++ b/extraconfig/tasks/major_upgrade_controller_pacemaker_2.sh @@ -50,6 +50,7 @@ mysql_need_update if [[ -n $(is_bootstrap_node) ]]; then if [ $DO_MYSQL_UPGRADE -eq 1 ]; then + backup_flags="--defaults-extra-file=/root/.my.cnf -u root --flush-privileges --all-databases --single-transaction" mysqldump $backup_flags > "$MYSQL_BACKUP_DIR/openstack_database.sql" cp -rdp /etc/my.cnf* "$MYSQL_BACKUP_DIR" fi diff --git a/extraconfig/tasks/major_upgrade_pacemaker.yaml b/extraconfig/tasks/major_upgrade_pacemaker.yaml index 8c91027d..74d3be71 100644 --- a/extraconfig/tasks/major_upgrade_pacemaker.yaml +++ b/extraconfig/tasks/major_upgrade_pacemaker.yaml @@ -18,10 +18,6 @@ parameters: constraints: - allowed_values: ['auto', 'yes', 'no'] default: 'auto' - IgnoreCephUpgradeWarnings: - type: boolean - default: false - description: If enabled, Ceph upgrade will be forced even though cluster or PGs status is not clean KeepSaharaServicesOnUpgrade: type: boolean default: true diff --git a/puppet/services/README.rst b/puppet/services/README.rst index e5c11535..223c3ed0 100644 --- a/puppet/services/README.rst +++ b/puppet/services/README.rst @@ -104,7 +104,9 @@ step, "step2" for the second, etc. 2) Stop all control-plane services, ready for upgrade - 3) Perform a package update, (either specific packages or the whole system) + 3) Perform a package update and install new packages: A general + upgrade is done, and only new package should go into service + ansible tasks. 4) Start services needed for migration tasks (e.g DB) diff --git a/puppet/services/ceph-mon.yaml b/puppet/services/ceph-mon.yaml index 1ce58335..d589ef89 100644 --- a/puppet/services/ceph-mon.yaml +++ b/puppet/services/ceph-mon.yaml @@ -59,6 +59,14 @@ parameters: } default: {} type: json + CephValidationRetries: + type: number + default: 5 + description: Number of retry attempts for Ceph validation + CephValidationDelay: + type: number + default: 10 + description: Interval (in seconds) in between validation checks MonitoringSubscriptionCephMon: default: 'overcloud-ceph-mon' type: string @@ -119,21 +127,32 @@ outputs: # rolling upgrade of all osd nodes in step1 - name: Check status tags: step0,validation - shell: ceph health | grep -qv HEALTH_ERR - # FIXME(shardy) I suspect we can use heat or ansible facts here instead? - - name: Get hostname + shell: ceph health | egrep -sq "HEALTH_OK|HEALTH_WARN" + - name: Stop CephMon tags: step0 - shell: hostname -s - register: mon_id - - name: Stop Ceph Mon + service: + name: ceph-mon@{{ ansible_hostname }} + state: stopped + - name: Update Ceph packages tags: step0 - service: name=ceph-mon@{{mon_id.stdout}} pattern=ceph-mon state=stopped - - name: Update ceph packages + yum: + name: ceph-mon + state: latest + - name: Start CephMon tags: step0 - yum: name=ceph-mon state=latest - - name: Start ceph-mon service - tags: step0 - service: name=ceph-mon@{{mon_id.stdout}} state=started + service: + name: ceph-mon@{{ ansible_hostname }} + state: started + # ceph-ansible + # https://github.com/ceph/ceph-ansible/blob/master/infrastructure-playbooks/rolling_update.yml#L149-L157 + - name: Wait for the monitor to join the quorum... + tags: step0,ceph_quorum_validation + shell: | + ceph -s | grep monmap | sed 's/.*quorum//' | egrep -sq {{ ansible_hostname }} + register: ceph_quorum_nodecheck + until: ceph_quorum_nodecheck.rc == 0 + retries: {get_param: CephValidationRetries} + delay: {get_param: CephValidationDelay} - name: ceph osd crush tunables default tags: step0 shell: ceph osd crush tunables default diff --git a/puppet/services/ceph-osd.yaml b/puppet/services/ceph-osd.yaml index 9bd83aab..a97fa116 100644 --- a/puppet/services/ceph-osd.yaml +++ b/puppet/services/ceph-osd.yaml @@ -21,6 +21,24 @@ parameters: MonitoringSubscriptionCephOsd: default: 'overcloud-ceph-osd' type: string + CephValidationRetries: + type: number + default: 40 + description: Number of retry attempts for Ceph validation + CephValidationDelay: + type: number + default: 30 + description: Interval (in seconds) in between validation checks + IgnoreCephUpgradeWarnings: + type: boolean + default: false + description: If enabled, Ceph upgrade will be forced even though cluster or PGs status is not clean + +parameter_groups: +- label: deprecated + description: Do not use deprecated params, they will be removed. + parameters: + - IgnoreCephUpgradeWarnings resources: CephBase: @@ -66,17 +84,37 @@ outputs: - name: ceph osd set noscrub tags: step1 command: ceph osd set noscrub - - name: Stop Ceph OSD + - name: Stop CephOSD tags: step1 - service: name=ceph-osd@{{ item }} state=stopped + service: + name: ceph-osd@{{ item }} + state: stopped with_items: "{{osd_ids.stdout.strip().split()}}" - - name: Update ceph OSD packages + - name: Update Ceph packages tags: step1 - yum: name=ceph-osd state=latest - - name: Start ceph-osd service + yum: + name: ceph-osd + state: latest + - name: Start CephOSD tags: step1 - service: name=ceph-osd@{{ item }} state=started + service: + name: ceph-osd@{{ item }} + state: started with_items: "{{osd_ids.stdout.strip().split()}}" + # with awk we are meant to check if $2 and $4 are *the same* but it returns 1 when + # they are, so the check is inverted to produce an useful exit code + - name: Wait for clean pgs... + tags: step1,ceph_pgs_clean_validation + vars: + ignore_warnings: {get_param: IgnoreCephUpgradeWarnings} + shell: | + ceph pg stat | awk '{exit($2!=$4)}' && ceph health | egrep -sq "HEALTH_OK|HEALTH_WARN" + register: ceph_pgs_healthcheck + until: ceph_pgs_healthcheck.rc == 0 + retries: {get_param: CephValidationRetries} + delay: {get_param: CephValidationDelay} + when: + - not ignore_warnings - name: ceph osd unset noout tags: step1 command: ceph osd unset noout diff --git a/puppet/services/ceph-rgw.yaml b/puppet/services/ceph-rgw.yaml index d7014e54..c5b29c7e 100644 --- a/puppet/services/ceph-rgw.yaml +++ b/puppet/services/ceph-rgw.yaml @@ -87,4 +87,6 @@ outputs: tags: step0,validation - name: Stop RGW instance tags: step1 - service: name=ceph-radosgw@{{rgw_id.stdout}} state=stopped + service: + name: ceph-radosgw@{{rgw_id.stdout}} + state: stopped diff --git a/puppet/services/nova-api.yaml b/puppet/services/nova-api.yaml index 0adefecd..f27b53f2 100644 --- a/puppet/services/nova-api.yaml +++ b/puppet/services/nova-api.yaml @@ -197,9 +197,6 @@ outputs: - name: Stop and disable nova_api service (pre-upgrade not under httpd) tags: step2 service: name=openstack-nova-api state=stopped enabled=no - - name: update nova api - tags: step2 - yum: name=openstack-nova-api state=latest - name: Create puppet manifest to set transport_url in nova.conf tags: step5 when: is_bootstrap_node diff --git a/puppet/services/nova-conductor.yaml b/puppet/services/nova-conductor.yaml index 7b086536..a19d0f8d 100644 --- a/puppet/services/nova-conductor.yaml +++ b/puppet/services/nova-conductor.yaml @@ -69,13 +69,10 @@ outputs: - name: Stop nova_conductor service tags: step2 service: name=openstack-nova-conductor state=stopped - - name: update nova conductor - tags: step2 - yum: name=openstack-nova-conductor state=latest # If not already set by puppet (e.g a pre-ocata version), set the # upgrade_level for compute to "auto" - name: Set compute upgrade level to auto - tags: step3 + tags: step2 ini_file: str_replace: template: "dest=/etc/nova/nova.conf section=upgrade_levels option=compute value=LEVEL" diff --git a/puppet/services/nova-scheduler.yaml b/puppet/services/nova-scheduler.yaml index 0e0b9d1e..e08bf182 100644 --- a/puppet/services/nova-scheduler.yaml +++ b/puppet/services/nova-scheduler.yaml @@ -67,6 +67,3 @@ outputs: - name: Stop nova_scheduler service tags: step2 service: name=openstack-nova-scheduler state=stopped - - name: update nova scheduler - tags: step2 - yum: name=openstack-nova-scheduler state=latest |