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     if ! pcs resource defaults | grep "resource-stickiness: INFINITY"; then
  78         pcs -f $pacemaker_dumpfile resource defaults resource-stickiness=INFINITY
  79     fi
  80
  81     echo "Setting resource start/stop timeouts"
  82     SERVICES="
  83 haproxy
  84 memcached
  85 httpd
  86 neutron-dhcp-agent
  87 neutron-l3-agent
  88 neutron-metadata-agent
  89 neutron-openvswitch-agent
  90 neutron-server
  91 openstack-ceilometer-alarm-evaluator
  92 openstack-ceilometer-alarm-notifier
  93 openstack-ceilometer-api
  94 openstack-ceilometer-central
  95 openstack-ceilometer-collector
  96 openstack-ceilometer-notification
  97 openstack-cinder-api
  98 openstack-cinder-scheduler
  99 openstack-cinder-volume
 100 openstack-glance-api
 101 openstack-glance-registry
 102 openstack-heat-api
 103 openstack-heat-api-cfn
 104 openstack-heat-api-cloudwatch
 105 openstack-heat-engine
 106 openstack-keystone
 107 openstack-nova-api
 108 openstack-nova-conductor
 109 openstack-nova-consoleauth
 110 openstack-nova-novncproxy
 111 openstack-nova-scheduler"
 112     for service in $SERVICES; do
 113         pcs -f $pacemaker_dumpfile resource update $service op start timeout=100s op stop timeout=100s
 114     done
 115     # mongod start timeout is higher, setting only stop timeout
 116     pcs resource update mongod op stop timeout=100s
 117
 118     echo "Applying new Pacemaker config"
 119     pcs cluster cib-push $pacemaker_dumpfile
 120
 121     echo "Pacemaker running, stopping cluster node and doing full package update"
 122     node_count=$(pcs status xml | grep -o "<nodes_configured.*/>" | grep -o 'number="[0-9]*"' | grep -o "[0-9]*")
 123     if [[ "$node_count" == "1" ]] ; then
 124         echo "Active node count is 1, stopping node with --force"
 125         pcs cluster stop --force
 126     else
 127         pcs cluster stop
 128     fi
 129 else
 130     echo "Excluding upgrading packages that are handled by config management tooling"
 131     command_arguments="$command_arguments --skip-broken"
 132     for exclude in $(cat /var/lib/tripleo/installed-packages/* | sort -u); do
 133         command_arguments="$command_arguments --exclude $exclude"
 134     done
 135 fi
 136
 137 command=${command:-update}
 138 full_command="yum -y $command $command_arguments"
 139 echo "Running: $full_command"
 140
 141 result=$($full_command)
 142 return_code=$?
 143 echo "$result"
 144 echo "yum return code: $return_code"
 145
 146 if [[ "$pacemaker_status" == "active" ]] ; then
 147     echo "Starting cluster node"
 148     pcs cluster start
 149
 150     hostname=$(hostname -s)
 151     tstart=$(date +%s)
 152     while [[ "$(pcs status | grep "^Online" | grep -F -o $hostname)" == "" ]]; do
 153         sleep 5
 154         tnow=$(date +%s)
 155         if (( tnow-tstart > cluster_start_timeout )) ; then
 156             echo "ERROR $hostname failed to join cluster in $cluster_start_timeout seconds"
 157             pcs status
 158             exit 1
 159         fi
 160     done
 161
 162     tstart=$(date +%s)
 163     while ! clustercheck; do
 164         sleep 5
 165         tnow=$(date +%s)
 166         if (( tnow-tstart > galera_sync_timeout )) ; then
 167             echo "ERROR galera sync timed out"
 168             exit 1
 169         fi
 170     done
 171
 172     pcs status
 173
 174 else
 175     echo -n "true" > $heat_outputs_path.update_managed_packages
 176 fi
 177
 178 echo "Finished yum_update.sh on server $deploy_server_id at `date`"
 179
 180 exit $return_code