extraconfig/tasks/yum_update.sh

   1 #!/bin/bash
   2
   3 # A heat-config-script which runs yum update during a stack-update.
   4 # Inputs:
   5 #   deploy_action - yum will only be run if this is UPDATE
   6 #   update_identifier - yum will only run for previously unused values of update_identifier
   7 #   command - yum sub-command to run, defaults to "update"
   8 #   command_arguments - yum command arguments, defaults to ""
   9
  10 echo "Started yum_update.sh on server $deploy_server_id at `date`"
  11 echo -n "false" > $heat_outputs_path.update_managed_packages
  12
  13 if [[ -z "$update_identifier" ]]; then
  14     echo "Not running due to unset update_identifier"
  15     exit 0
  16 fi
  17
  18 timestamp_dir=/var/lib/overcloud-yum-update
  19 mkdir -p $timestamp_dir
  20
  21 # sanitise to remove unusual characters
  22 update_identifier=${update_identifier//[^a-zA-Z0-9-_]/}
  23
  24 # seconds to wait for this node to rejoin the cluster after update
  25 cluster_start_timeout=600
  26 galera_sync_timeout=360
  27
  28 timestamp_file="$timestamp_dir/$update_identifier"
  29 if [[ -a "$timestamp_file" ]]; then
  30     echo "Not running for already-run timestamp \"$update_identifier\""
  31     exit 0
  32 fi
  33 touch "$timestamp_file"
  34
  35 command_arguments=${command_arguments:-}
  36
  37 list_updates=$(yum list updates)
  38
  39 if [[ "$list_updates" == "" ]]; then
  40     echo "No packages require updating"
  41     exit 0
  42 fi
  43
  44 pacemaker_status=$(systemctl is-active pacemaker)
  45 pacemaker_dumpfile=$(mktemp)
  46
  47 if [[ "$pacemaker_status" == "active" ]] ; then
  48     echo "Dumping Pacemaker config"
  49     pcs cluster cib $pacemaker_dumpfile
  50
  51     echo "Checking for missing constraints"
  52
  53     if ! pcs constraint order show | grep "start openstack-nova-novncproxy-clone then start openstack-nova-api-clone"; then
  54         pcs -f $pacemaker_dumpfile constraint order start openstack-nova-novncproxy-clone then openstack-nova-api-clone
  55     fi
  56
  57     if ! pcs constraint order show | grep "start rabbitmq-clone then start openstack-keystone-clone"; then
  58         pcs -f $pacemaker_dumpfile constraint order start rabbitmq-clone then openstack-keystone-clone
  59     fi
  60
  61     if ! pcs constraint order show | grep "promote galera-master then start openstack-keystone-clone"; then
  62         pcs -f $pacemaker_dumpfile constraint order promote galera-master then openstack-keystone-clone
  63     fi
  64
  65     if ! pcs constraint order show | grep "start haproxy-clone then start openstack-keystone-clone"; then
  66         pcs -f $pacemaker_dumpfile constraint order start haproxy-clone then openstack-keystone-clone
  67     fi
  68
  69     if ! pcs constraint order show | grep "start memcached-clone then start openstack-keystone-clone"; then
  70         pcs -f $pacemaker_dumpfile constraint order start memcached-clone then openstack-keystone-clone
  71     fi
  72
  73     if ! pcs constraint order show | grep "promote redis-master then start openstack-ceilometer-central-clone"; then
  74         pcs -f $pacemaker_dumpfile constraint order promote redis-master then start openstack-ceilometer-central-clone require-all=false
  75     fi
  76
  77     # ensure neutron constraints https://review.openstack.org/#/c/229466
  78     # remove ovs-cleanup after server and add openvswitch-agent instead
  79     if  pcs constraint order show  | grep "start neutron-server-clone then start neutron-ovs-cleanup-clone"; then
  80         pcs -f $pacemaker_dumpfile constraint remove order-neutron-server-clone-neutron-ovs-cleanup-clone-mandatory
  81     fi
  82     if ! pcs constraint order show | grep "start neutron-server-clone then start neutron-openvswitch-agent-clone"; then
  83         pcs -f $pacemaker_dumpfile constraint order start neutron-server-clone then neutron-openvswitch-agent-clone
  84     fi
  85
  86
  87     if ! pcs resource defaults | grep "resource-stickiness: INFINITY"; then
  88         pcs -f $pacemaker_dumpfile resource defaults resource-stickiness=INFINITY
  89     fi
  90
  91     echo "Setting resource start/stop timeouts"
  92     SERVICES="
  93 haproxy
  94 memcached
  95 httpd
  96 neutron-dhcp-agent
  97 neutron-l3-agent
  98 neutron-metadata-agent
  99 neutron-openvswitch-agent
 100 neutron-server
 101 openstack-ceilometer-alarm-evaluator
 102 openstack-ceilometer-alarm-notifier
 103 openstack-ceilometer-api
 104 openstack-ceilometer-central
 105 openstack-ceilometer-collector
 106 openstack-ceilometer-notification
 107 openstack-cinder-api
 108 openstack-cinder-scheduler
 109 openstack-cinder-volume
 110 openstack-glance-api
 111 openstack-glance-registry
 112 openstack-heat-api
 113 openstack-heat-api-cfn
 114 openstack-heat-api-cloudwatch
 115 openstack-heat-engine
 116 openstack-keystone
 117 openstack-nova-api
 118 openstack-nova-conductor
 119 openstack-nova-consoleauth
 120 openstack-nova-novncproxy
 121 openstack-nova-scheduler"
 122     for service in $SERVICES; do
 123         pcs -f $pacemaker_dumpfile resource update $service op start timeout=100s op stop timeout=100s
 124     done
 125     # mongod start timeout is higher, setting only stop timeout
 126     pcs resource update mongod op stop timeout=100s
 127
 128     echo "Applying new Pacemaker config"
 129     pcs cluster cib-push $pacemaker_dumpfile
 130
 131     echo "Pacemaker running, stopping cluster node and doing full package update"
 132     node_count=$(pcs status xml | grep -o "<nodes_configured.*/>" | grep -o 'number="[0-9]*"' | grep -o "[0-9]*")
 133     if [[ "$node_count" == "1" ]] ; then
 134         echo "Active node count is 1, stopping node with --force"
 135         pcs cluster stop --force
 136     else
 137         pcs cluster stop
 138     fi
 139 else
 140     echo "Excluding upgrading packages that are handled by config management tooling"
 141     command_arguments="$command_arguments --skip-broken"
 142     for exclude in $(cat /var/lib/tripleo/installed-packages/* | sort -u); do
 143         command_arguments="$command_arguments --exclude $exclude"
 144     done
 145 fi
 146
 147 command=${command:-update}
 148 full_command="yum -y $command $command_arguments"
 149 echo "Running: $full_command"
 150
 151 result=$($full_command)
 152 return_code=$?
 153 echo "$result"
 154 echo "yum return code: $return_code"
 155
 156 if [[ "$pacemaker_status" == "active" ]] ; then
 157     echo "Starting cluster node"
 158     pcs cluster start
 159
 160     hostname=$(hostname -s)
 161     tstart=$(date +%s)
 162     while [[ "$(pcs status | grep "^Online" | grep -F -o $hostname)" == "" ]]; do
 163         sleep 5
 164         tnow=$(date +%s)
 165         if (( tnow-tstart > cluster_start_timeout )) ; then
 166             echo "ERROR $hostname failed to join cluster in $cluster_start_timeout seconds"
 167             pcs status
 168             exit 1
 169         fi
 170     done
 171
 172     tstart=$(date +%s)
 173     while ! clustercheck; do
 174         sleep 5
 175         tnow=$(date +%s)
 176         if (( tnow-tstart > galera_sync_timeout )) ; then
 177             echo "ERROR galera sync timed out"
 178             exit 1
 179         fi
 180     done
 181
 182     pcs status
 183
 184 else
 185     echo -n "true" > $heat_outputs_path.update_managed_packages
 186 fi
 187
 188 echo "Finished yum_update.sh on server $deploy_server_id at `date`"
 189
 190 exit $return_code