5 cluster_sync_timeout=1800
7 # After migrating the cluster to HA-NG the services not under pacemaker's control
8 # are still up and running. We need to stop them explicitely otherwise during the yum
9 # upgrade the rpm %post sections will try to do a systemctl try-restart <service>, which
10 # is going to take a long time because rabbit is down. By having the service stopped
11 # systemctl try-restart is a noop
13 for service in $(services_to_migrate); do
14 manage_systemd_service stop "${service%%-clone}"
15 # So the reason for not reusing check_resource_systemd is that
16 # I have observed systemctl is-active returning unknown with at least
17 # one service that was stopped (See LP 1627254)
20 tend=$(( $tstart + $timeout ))
22 while (( $(date +%s) < $tend )); do
23 if [[ "$(systemctl is-active ${service%%-clone})" = "active" ]]; then
24 echo "$service still active, sleeping $check_interval seconds."
27 # we do not care if it is inactive, unknown or failed as long as it is
35 # In case the mysql package is updated, the database on disk must be
36 # upgraded as well. This typically needs to happen during major
37 # version upgrades (e.g. 5.5 -> 5.6, 5.5 -> 10.1...)
39 # Because in-place upgrades are not supported across 2+ major versions
40 # (e.g. 5.5 -> 10.1), we rely on logical upgrades via dump/restore cycle
41 # https://bugzilla.redhat.com/show_bug.cgi?id=1341968
43 # The default is to determine automatically if upgrade is needed based
44 # on mysql package versionning, but this can be overriden manually
45 # to support specific upgrade scenario
47 # Calling this function will set the DO_MYSQL_UPGRADE variable which is used
51 if [[ -n $(is_bootstrap_node) ]]; then
52 if [ $DO_MYSQL_UPGRADE -eq 1 ]; then
53 mysqldump $backup_flags > "$MYSQL_BACKUP_DIR/openstack_database.sql"
54 cp -rdp /etc/my.cnf* "$MYSQL_BACKUP_DIR"
57 pcs resource disable redis
58 check_resource redis stopped 600
59 pcs resource disable rabbitmq
60 check_resource rabbitmq stopped 600
61 pcs resource disable galera
62 check_resource galera stopped 600
63 pcs resource disable openstack-cinder-volume
64 check_resource openstack-cinder-volume stopped 600
65 # Disable all VIPs before stopping the cluster, so that pcs doesn't use one as a source address:
66 # https://bugzilla.redhat.com/show_bug.cgi?id=1330688
67 for vip in $(pcs resource show | grep ocf::heartbeat:IPaddr2 | grep Started | awk '{ print $1 }'); do
68 pcs resource disable $vip
69 check_resource $vip stopped 60
71 pcs cluster stop --all
75 # Swift isn't controlled by pacemaker
79 while systemctl is-active pacemaker; do
82 if (( tnow-tstart > cluster_sync_timeout )) ; then
83 echo_error "ERROR: cluster shutdown timed out"
88 # The reason we do an sql dump *and* we move the old dir out of
89 # the way is because it gives us an extra level of safety in case
90 # something goes wrong during the upgrade. Once the restore is
91 # successful we go ahead and remove it. If the directory exists
92 # we bail out as it means the upgrade process had issues in the last
94 if [ $DO_MYSQL_UPGRADE -eq 1 ]; then
95 if [ -d $MYSQL_TEMP_UPGRADE_BACKUP_DIR ]; then
96 echo_error "ERROR: mysql backup dir already exist"
99 mv /var/lib/mysql $MYSQL_TEMP_UPGRADE_BACKUP_DIR
102 # Special-case OVS for https://bugs.launchpad.net/tripleo/+bug/1635205
103 if [[ -n $(rpm -q --scripts openvswitch | awk '/postuninstall/,/*/' | grep "systemctl.*try-restart") ]]; then
104 echo "Manual upgrade of openvswitch - restart in postun detected"
105 mkdir OVS_UPGRADE || true
107 echo "Attempting to downloading latest openvswitch with yumdownloader"
108 yumdownloader --resolve openvswitch
109 echo "Updating openvswitch with nopostun option"
110 rpm -U --replacepkgs --nopostun ./*.rpm
113 echo "Skipping manual upgrade of openvswitch - no restart in postun detected"
116 yum -y install python-zaqarclient # needed for os-collect-config
119 # We need to ensure at least those two configuration settings, otherwise
120 # mariadb 10.1+ won't activate galera replication.
121 # wsrep_cluster_address must only be set though, its value does not
122 # matter because it's overriden by the galera resource agent.
123 cat >> /etc/my.cnf.d/galera.cnf <<EOF
126 wsrep_cluster_address = gcomm://localhost
129 if [ $DO_MYSQL_UPGRADE -eq 1 ]; then
130 # Scripts run via heat have no HOME variable set and this confuses
134 mkdir /var/lib/mysql || /bin/true
135 chown mysql:mysql /var/lib/mysql
136 chmod 0755 /var/lib/mysql
137 restorecon -R /var/lib/mysql/
138 mysql_install_db --datadir=/var/lib/mysql --user=mysql
139 chown -R mysql:mysql /var/lib/mysql/
141 if [ "$(hiera -c /etc/puppet/hiera.yaml bootstrap_nodeid)" = "$(facter hostname)" ]; then
142 mysqld_safe --wsrep-new-cluster &
143 # We have a populated /root/.my.cnf with root/password here so
144 # we need to temporarily rename it because the newly created
145 # db is empty and no root password is set
146 mv /root/.my.cnf /root/.my.cnf.temporary
147 timeout 60 sh -c 'while ! mysql -e "" &> /dev/null; do sleep 1; done'
148 mysql -u root < "$MYSQL_BACKUP_DIR/openstack_database.sql"
149 mv /root/.my.cnf.temporary /root/.my.cnf
150 mysqladmin -u root shutdown
151 # The import was successful so we may remove the folder
152 rm -r "$MYSQL_BACKUP_DIR"
156 # If we reached here without error we can safely blow away the origin
157 # mysql dir from every controller
159 # TODO: What if the upgrade fails on the bootstrap node, but not on
160 # this controller. Data may be lost.
161 if [ $DO_MYSQL_UPGRADE -eq 1 ]; then
162 rm -r $MYSQL_TEMP_UPGRADE_BACKUP_DIR
165 # Let's reset the stonith back to true if it was true, before starting the cluster
166 if [[ -n $(is_bootstrap_node) ]]; then
167 if [ -f /var/tmp/stonith-true ]; then
168 pcs -f /var/lib/pacemaker/cib/cib.xml property set stonith-enabled=true
170 rm -f /var/tmp/stonith-true
173 # Pin messages sent to compute nodes to kilo, these will be upgraded later
174 crudini --set /etc/nova/nova.conf upgrade_levels compute "$upgrade_level_nova_compute"
175 # https://bugzilla.redhat.com/show_bug.cgi?id=1284047
176 # Change-Id: Ib3f6c12ff5471e1f017f28b16b1e6496a4a4b435
177 crudini --set /etc/ceilometer/ceilometer.conf DEFAULT rpc_backend rabbit
178 # https://bugzilla.redhat.com/show_bug.cgi?id=1284058
179 # Ifd1861e3df46fad0e44ff9b5cbd58711bbc87c97 Swift Ceilometer middleware no longer exists
180 crudini --set /etc/swift/proxy-server.conf pipeline:main pipeline "catch_errors healthcheck cache ratelimit tempurl formpost authtoken keystone staticweb proxy-logging proxy-server"
181 # LP: 1615035, required only for M/N upgrade.
182 crudini --set /etc/nova/nova.conf DEFAULT scheduler_host_manager host_manager
183 # LP: 1627450, required only for M/N upgrade
184 crudini --set /etc/nova/nova.conf DEFAULT scheduler_driver filter_scheduler
186 crudini --set /etc/sahara/sahara.conf DEFAULT plugins ambari,cdh,mapr,vanilla,spark,storm