diff --git a/heartbeat/conntrackd.in b/heartbeat/conntrackd.in index 1c2ee955b..87989f4aa 100644 --- a/heartbeat/conntrackd.in +++ b/heartbeat/conntrackd.in @@ -1,335 +1,335 @@ #!@BASH_SHELL@ # # # An OCF RA for conntrackd # http://conntrack-tools.netfilter.org/ # # Copyright (c) 2011 Dominik Klein # # This program is free software; you can redistribute it and/or modify # it under the terms of version 2 of the GNU General Public License as # published by the Free Software Foundation. # # This program is distributed in the hope that it would be useful, but # WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. # # Further, this software is distributed without any warranty that it is # free of the rightful claim of any third person regarding infringement # or the like. Any license provided herein, whether implied or # otherwise, applies only to this software file. Patent licenses, if # any, provided herein do not apply to combinations of this program with # other software, or any other product whatsoever. # # You should have received a copy of the GNU General Public License # along with this program; if not, write the Free Software Foundation, # Inc., 59 Temple Place - Suite 330, Boston MA 02111-1307, USA. # ####################################################################### # Initialization: : ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat} . ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs ####################################################################### OCF_RESKEY_binary_default=conntrackd OCF_RESKEY_config_default=/etc/conntrackd/conntrackd.conf # For users of versions prior to 1.2: # Map renamed parameter "conntrackd" to "binary" if in use : ${OCF_RESKEY_binary=${OCF_RESKEY_conntrackd-${OCF_RESKEY_binary_default}}} : ${OCF_RESKEY_config=${OCF_RESKEY_config_default}} meta_data() { cat < 1.0 Promotable OCF Resource Agent for conntrackd This resource agent manages conntrackd Name of the conntrackd executable. If conntrackd is installed and available in the default PATH, it is sufficient to configure the name of the binary For example "my-conntrackd-binary-version-0.9.14" If conntrackd is installed somewhere else, you may also give a full path For example "/packages/conntrackd-0.9.14/sbin/conntrackd" Name of the conntrackd executable Full path to the conntrackd.conf file. For example "/packages/conntrackd-0.9.14/etc/conntrackd/conntrackd.conf" Path to conntrackd.conf END } meta_expect_eq() { local what=$1 whatvar=OCF_RESKEY_CRM_meta_${1//-/_} expect=$2 local val=${!whatvar} if [[ -n $val ]]; then # [, not [[, or it won't work ;) [ $val = $expect ] && return fi ocf_exit_reason "meta parameter misconfigured, expected $what $op $expect, but found ${val:-unset}." exit $OCF_ERR_CONFIGURED } conntrackd_is_master() { # You can't query conntrackd whether it is master or slave. It can be both at the same time. # This RA creates a statefile during promote and enforces master-max=1 and clone-node-max=1 ha_pseudo_resource $statefile monitor } conntrackd_set_master_score() { - ${HA_SBIN_DIR}/crm_master -Q -l reboot -v $1 + ${HA_SBIN_DIR}/crm_master -q -l reboot -v $1 } conntrackd_monitor() { rc=$OCF_NOT_RUNNING # It does not write a PID file, so check the socket exists after # extracting its path from the configuration file local conntrack_socket=$(awk '/^[ \t]*UNIX[ \t]*{/,/^[ \t]*}/ { if ($1 == "Path") { print $2 } }' $OCF_RESKEY_config) [ -S "$conntrack_socket" ] && rc=$OCF_SUCCESS if [ "$rc" -eq "$OCF_SUCCESS" ]; then # conntrackd is running # now see if it acceppts queries if ! $OCF_RESKEY_binary -C $OCF_RESKEY_config -s > /dev/null 2>&1; then rc=$OCF_ERR_GENERIC ocf_exit_reason "conntrackd is running but not responding to queries" fi if conntrackd_is_master; then rc=$OCF_RUNNING_MASTER # Restore master setting on probes if [ $OCF_RESKEY_CRM_meta_interval -eq 0 ]; then conntrackd_set_master_score $master_score fi else # Restore master setting on probes if [ $OCF_RESKEY_CRM_meta_interval -eq 0 ]; then conntrackd_set_master_score $slave_score fi fi fi return $rc } conntrackd_start() { rc=$OCF_ERR_GENERIC # Keep trying to start the resource; # wait for the CRM to time us out if this fails while :; do conntrackd_monitor status=$? case "$status" in $OCF_SUCCESS) conntrackd_set_master_score $slave_score # -n = request resync from the others if ! $OCF_RESKEY_binary -C $OCF_RESKEY_config -n; then ocf_exit_reason "$OCF_RESKEY_binary -C $OCF_RESKEY_config -n failed during start." rc=$OCF_ERR_GENERIC else rc=$OCF_SUCCESS fi break ;; $OCF_NOT_RUNNING) ocf_log info "Starting conntrackd" $OCF_RESKEY_binary -C $OCF_RESKEY_config -d ;; $OCF_RUNNING_MASTER) ocf_log warn "conntrackd already in master mode, demoting." ha_pseudo_resource $statefile stop ;; $OCF_ERR_GENERIC) ocf_exit_reason "conntrackd start failed" rc=$OCF_ERR_GENERIC break ;; esac done return $rc } conntrackd_stop() { rc=$OCF_ERR_GENERIC # Keep trying to bring down the resource; # wait for the CRM to time us out if this fails while :; do conntrackd_monitor status=$? case "$status" in $OCF_SUCCESS|$OCF_ERR_GENERIC) ocf_log info "Stopping conntrackd" $OCF_RESKEY_binary -C $OCF_RESKEY_config -k ;; $OCF_NOT_RUNNING) rc=$OCF_SUCCESS break ;; $OCF_RUNNING_MASTER) ocf_log warn "conntrackd still master" ;; esac done return $rc } conntrackd_validate_all() { check_binary "$OCF_RESKEY_binary" if ! [ -e "$OCF_RESKEY_config" ]; then ocf_exit_reason "Config FILE $OCF_RESKEY_config does not exist" return $OCF_ERR_INSTALLED fi meta_expect_eq master-node-max 1 meta_expect_eq master-max 1 meta_expect_eq clone-node-max 1 return $OCF_SUCCESS } conntrackd_promote() { rc=$OCF_SUCCESS if ! conntrackd_is_master; then # -c = Commit the external cache to the kernel # -f = Flush internal and external cache # -R = resync with the kernel table # -B = send a bulk update on the line for parm in c f R B; do if ! $OCF_RESKEY_binary -C $OCF_RESKEY_config -$parm; then ocf_exit_reason "$OCF_RESKEY_binary -C $OCF_RESKEY_config -$parm failed during promote." rc=$OCF_ERR_GENERIC break fi done ha_pseudo_resource $statefile start conntrackd_set_master_score $master_score fi return $rc } conntrackd_demote() { rc=$OCF_SUCCESS if conntrackd_is_master; then # -t = shorten kernel timers to remove zombies # -n = request a resync from the others for parm in t n; do if ! $OCF_RESKEY_binary -C $OCF_RESKEY_config -$parm; then ocf_exit_reason "$OCF_RESKEY_binary -C $OCF_RESKEY_config -$parm failed during demote." rc=$OCF_ERR_GENERIC break fi done ha_pseudo_resource $statefile stop conntrackd_set_master_score $slave_score fi return $rc } conntrackd_notify() { hostname=$(hostname) # OCF_RESKEY_CRM_meta_notify_master_uname is a whitespace separated list of master hostnames for master in $OCF_RESKEY_CRM_meta_notify_master_uname; do # if we are the master and an instance was just started on another node: # send a bulk update to allow failback if [ "$hostname" = "$master" -a "$OCF_RESKEY_CRM_meta_notify_type" = "post" -a "$OCF_RESKEY_CRM_meta_notify_operation" = "start" -a "$OCF_RESKEY_CRM_meta_notify_start_uname" != "$hostname" ]; then ocf_log info "Sending bulk update in post start to peers to allow failback" $OCF_RESKEY_binary -C $OCF_RESKEY_config -B fi done for tobepromoted in $OCF_RESKEY_CRM_meta_notify_promote_uname; do # if there is a promote action to be executed on another node: # send a bulk update to allow failback if [ "$hostname" != "$tobepromoted" -a "$OCF_RESKEY_CRM_meta_notify_type" = "pre" -a "$OCF_RESKEY_CRM_meta_notify_operation" = "promote" ]; then ocf_log info "Sending bulk update in pre promote to peers to allow failback" $OCF_RESKEY_binary -C $OCF_RESKEY_config -B fi done } conntrackd_usage() { cat < # All Rights Reserved. # # This program is free software; you can redistribute it and/or modify # it under the terms of version 2 of the GNU General Public License as # published by the Free Software Foundation. # # This program is distributed in the hope that it would be useful, but # WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. # # Further, this software is distributed without any warranty that it is # free of the rightful claim of any third person regarding infringement # or the like. Any license provided herein, whether implied or # otherwise, applies only to this software file. Patent licenses, if # any, provided herein do not apply to combinations of this program with # other software, or any other product whatsoever. # # You should have received a copy of the GNU General Public License # along with this program; if not, write the Free Software Foundation, # Inc., 59 Temple Place - Suite 330, Boston MA 02111-1307, USA. # ## # README. # # This agent only supports being configured as a multistate Promoted # resource. # # Unpromoted vs Promoted role: # # During the 'Unpromoted' role, galera instances are in read-only mode and # will not attempt to connect to the cluster. This role exists only as # a means to determine which galera instance is the most up-to-date. The # most up-to-date node will be used to bootstrap a galera cluster that # has no current members. # # The galera instances will only begin to be promoted to the Promoted role # once all the nodes in the 'wsrep_cluster_address' connection address # have entered read-only mode. At that point the node containing the # database that is most current will be promoted to Promoted. Once the first # Promoted instance bootstraps the galera cluster, the other nodes will be # promoted to Promoted as well. # # Example: Create a galera cluster using nodes rhel7-node1 rhel7-node2 rhel7-node3 # # pcs resource create db galera enable_creation=true \ # wsrep_cluster_address="gcomm://rhel7-auto1,rhel7-auto2,rhel7-auto3" meta promoted-max=3 --promoted # # By setting the 'enable_creation' option, the database will be automatically # generated at startup. The meta attribute 'promoted-max=3' means that all 3 # nodes listed in the wsrep_cluster_address list will be allowed to connect # to the galera cluster and perform replication. # # NOTE: If you have more nodes in the pacemaker cluster then you wish # to have in the galera cluster, make sure to use location contraints to prevent # pacemaker from attempting to place a galera instance on a node that is # not in the 'wsrep_cluster_address" list. # ## ####################################################################### # Initialization: : ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat} . ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs if [ "$__OCF_ACTION" != "meta-data" ]; then . ${OCF_FUNCTIONS_DIR}/mysql-common.sh NODENAME=$(ocf_attribute_target) fi # It is common for some galera instances to store # check user that can be used to query status # in this file if [ -f "/etc/sysconfig/clustercheck" ]; then . /etc/sysconfig/clustercheck elif [ -f "/etc/default/clustercheck" ]; then . /etc/default/clustercheck fi # Parameter defaults OCF_RESKEY_wsrep_cluster_address_default="" OCF_RESKEY_cluster_host_map_default="" OCF_RESKEY_check_user_default="" OCF_RESKEY_check_passwd_default="" OCF_RESKEY_two_node_mode_default="false" : ${OCF_RESKEY_wsrep_cluster_address=${OCF_RESKEY_wsrep_cluster_address_default}} : ${OCF_RESKEY_cluster_host_map=${OCF_RESKEY_cluster_host_map_default}} : ${OCF_RESKEY_check_user=${OCF_RESKEY_check_user_default}} : ${OCF_RESKEY_check_passwd=${OCF_RESKEY_check_passwd_default}} : ${OCF_RESKEY_two_node_mode=${OCF_RESKEY_two_node_mode_default}} ####################################################################### # Defaults: OCF_RESKEY_check_passwd_use_empty_default=0 : ${OCF_RESKEY_check_passwd_use_empty=${OCF_RESKEY_check_passwd_use_empty_default}} ####################################################################### usage() { cat < 1.0 Resource script for managing galera database. Manages a galera instance Location of the MySQL server binary MySQL server binary Location of the MySQL client binary MySQL client binary Configuration file MySQL config Directory containing databases MySQL datadir User running MySQL daemon MySQL user Group running MySQL daemon (for logfile and directory permissions) MySQL group The logfile to be used for mysqld. MySQL log file The pidfile to be used for mysqld. MySQL pid file The socket to be used for mysqld. MySQL socket If the MySQL database does not exist, it will be created Create the database if it does not exist Additional parameters which are passed to the mysqld on startup. (e.g. --skip-external-locking or --skip-grant-tables) Additional parameters to pass to mysqld The galera cluster address. This takes the form of: gcomm://node,node,node Only nodes present in this node list will be allowed to start a galera instance. The galera node names listed in this address are expected to match valid pacemaker node names. If both names need to differ, you must provide a mapping in option cluster_host_map. Galera cluster address A mapping of pacemaker node names to galera node names. To be used when both pacemaker and galera names need to differ, (e.g. when galera names map to IP from a specific network interface) This takes the form of: pcmk1:node.1.galera;pcmk2:node.2.galera;pcmk3:node.3.galera where the galera resource started on node pcmk1 would be named node.1.galera in the wsrep_cluster_address Pacemaker to Galera name mapping Cluster check user. MySQL test user Cluster check user password. Empty passwords are ignored unless the parameter "check_passwd_use_empty" is set to 1. check password Use an empty "check_passwd" password. If this parameter is set to 1, "check_passwd" will be ignored and an empty password is used when calling the "mysql" client binary. check password use empty If running in a 2-node pacemaker cluster, rely on pacemaker quorum to allow automatic recovery even when the other node is unreachable. Use it with caution! (and fencing) Special recovery when running on a 2-node cluster END } get_option_variable() { local key=$1 $MYSQL $MYSQL_OPTIONS_CHECK -e "SHOW VARIABLES like '$key';" | tail -1 } get_status_variable() { local key=$1 $MYSQL $MYSQL_OPTIONS_CHECK -e "show status like '$key';" | tail -1 } set_bootstrap_node() { local node=$(ocf_attribute_target $1) ${HA_SBIN_DIR}/crm_attribute -N $node -l reboot --name "${INSTANCE_ATTR_NAME}-bootstrap" -v "true" } clear_bootstrap_node() { ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-bootstrap" -D } is_bootstrap() { ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-bootstrap" --quiet 2>/dev/null } set_no_grastate() { ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-no-grastate" -v "true" } clear_no_grastate() { ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-no-grastate" -D } is_no_grastate() { local node=$(ocf_attribute_target $1) ${HA_SBIN_DIR}/crm_attribute -N $node -l reboot --name "${INSTANCE_ATTR_NAME}-no-grastate" --quiet 2>/dev/null } clear_last_commit() { ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-last-committed" -D } set_last_commit() { ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-last-committed" -v $1 } get_last_commit() { local node=$(ocf_attribute_target $1) if [ -z "$node" ]; then ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-last-committed" --quiet 2>/dev/null else ${HA_SBIN_DIR}/crm_attribute -N $node -l reboot --name "${INSTANCE_ATTR_NAME}-last-committed" --quiet 2>/dev/null fi } clear_safe_to_bootstrap() { ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-safe-to-bootstrap" -D } set_safe_to_bootstrap() { ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-safe-to-bootstrap" -v $1 } get_safe_to_bootstrap() { local node=$(ocf_attribute_target $1) if [ -z "$node" ]; then ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "${INSTANCE_ATTR_NAME}-safe-to-bootstrap" --quiet 2>/dev/null else ${HA_SBIN_DIR}/crm_attribute -N $node -l reboot --name "${INSTANCE_ATTR_NAME}-safe-to-bootstrap" --quiet 2>/dev/null fi } wait_for_sync() { local state=$(get_status_variable "wsrep_local_state") ocf_log info "Waiting for database to sync with the cluster. " while [ "$state" != "4" ]; do sleep 1 state=$(get_status_variable "wsrep_local_state") done ocf_log info "Database synced." } is_primary() { cluster_status=$(get_status_variable "wsrep_cluster_status") if [ "$cluster_status" = "Primary" ]; then return 0 fi if [ -z "$cluster_status" ]; then ocf_exit_reason "Unable to retrieve wsrep_cluster_status, verify check_user '$OCF_RESKEY_check_user' has permissions to view status" else ocf_log info "Galera instance wsrep_cluster_status=${cluster_status}" fi return 1 } is_readonly() { local res=$(get_option_variable "read_only") if ! ocf_is_true "$res"; then return 1 fi cluster_status=$(get_status_variable "wsrep_cluster_status") if ! [ "$cluster_status" = "Disconnected" ]; then return 1 fi return 0 } is_two_node_mode_active() { # crm_node or corosync-quorumtool cannot access various corosync # flags when running inside a bundle, so only count the cluster # members - ocf_is_true "$OCF_RESKEY_two_node_mode" && crm_mon_no_validation -1X | xmllint --xpath "count(//nodes/node[@type='member'])" - | grep -q -w 2 + ocf_is_true "$OCF_RESKEY_two_node_mode" && crm_mon_no_validation -1 $XMLOPT | xmllint --xpath "count(//nodes/node[@type='member'])" - | grep -q -w 2 } is_last_node_in_quorate_partition() { # when a network split occurs in a 2-node cluster, pacemaker # fences the other node and try to retain quorum. So until # the fencing is resolved (and the status of the peer node # is clean), we shouldn't consider ourself quorate. local partition_members=$(${HA_SBIN_DIR}/crm_node -p | wc -w) local quorate=$(${HA_SBIN_DIR}/crm_node -q) - local clean_members=$(crm_mon_no_validation -1X | xmllint --xpath 'count(//nodes/node[@type="member" and @unclean="false"])' -) + local clean_members=$(crm_mon_no_validation -1 $XMLOPT | xmllint --xpath 'count(//nodes/node[@type="member" and @unclean="false"])' -) [ "$partition_members" = 1 ] && [ "$quorate" = 1 ] && [ "$clean_members" = 2 ] } master_exists() { if [ "$__OCF_ACTION" = "demote" ]; then # We don't want to detect master instances during demote. # 1. we could be detecting ourselves as being master, which is no longer the case. # 2. we could be detecting other master instances that are in the process of shutting down. # by not detecting other master instances in "demote" we are deferring this check # to the next recurring monitor operation which will be much more accurate return 1 fi # determine if a master instance is already up and is healthy - ocf_version_cmp "$OCF_RESKEY_crm_feature_set" "3.1.0" - res=$? - if [ -z "$OCF_RESKEY_crm_feature_set" ] || [ $res -eq 2 ]; then - XMLOPT="--output-as=xml" - ocf_version_cmp "$OCF_RESKEY_crm_feature_set" "3.2.0" - if [ $? -eq 1 ]; then - crm_mon_no_validation -1 $XMLOPT >/dev/null 2>&1 - if [ $? -ne 0 ]; then - XMLOPT="--as-xml" - fi - fi - else - XMLOPT="--as-xml" - fi crm_mon_no_validation -1 $XMLOPT | grep -q -i -E "resource.*id=\"${INSTANCE_ATTR_NAME}\".*role=\"(Promoted|Master)\".*active=\"true\".*orphaned=\"false\".*failed=\"false\"" return $? } clear_master_score() { local node=$(ocf_attribute_target $1) if [ -z "$node" ]; then ocf_promotion_score -D else ocf_promotion_score -D -N $node fi } set_master_score() { local node=$(ocf_attribute_target $1) if [ -z "$node" ]; then ocf_promotion_score -v 100 else ocf_promotion_score -N $node -v 100 fi } promote_everyone() { for node in $(echo "$OCF_RESKEY_wsrep_cluster_address" | sed 's/gcomm:\/\///g' | tr -d ' ' | tr -s ',' ' '); do local pcmk_node=$(galera_to_pcmk_name $node) if [ -z "$pcmk_node" ]; then ocf_log err "Could not determine pacemaker node from galera name <${node}>." return else node=$pcmk_node fi set_master_score $node done } greater_than_equal_long() { # there are values we need to compare in this script # that are too large for shell -gt to process echo | awk -v n1="$1" -v n2="$2" '{if (n1>=n2) printf ("true"); else printf ("false");}' | grep -q "true" } galera_to_pcmk_name() { local galera=$1 if [ -z "$OCF_RESKEY_cluster_host_map" ]; then echo $galera else echo "$OCF_RESKEY_cluster_host_map" | tr ';' '\n' | tr -d ' ' | sed 's/:/ /' | awk -F' ' '$2=="'"$galera"'" {print $1;exit}' fi } pcmk_to_galera_name() { local pcmk=$1 if [ -z "$OCF_RESKEY_cluster_host_map" ]; then echo $pcmk else echo "$OCF_RESKEY_cluster_host_map" | tr ';' '\n' | tr -d ' ' | sed 's/:/ /' | awk -F' ' '$1=="'"$pcmk"'" {print $2;exit}' fi } detect_first_master() { local best_commit=0 local last_commit=0 local missing_nodes=0 local nodes="" local nodes_recovered="" local all_nodes local best_node_gcomm local best_node local safe_to_bootstrap all_nodes=$(echo "$OCF_RESKEY_wsrep_cluster_address" | sed 's/gcomm:\/\///g' | tr -d ' ' | tr -s ',' ' ') best_node_gcomm=$(echo "$all_nodes" | sed 's/^.* \(.*\)$/\1/') best_node=$(galera_to_pcmk_name $best_node_gcomm) if [ -z "$best_node" ]; then ocf_log err "Could not determine initial best node from galera name <${best_node_gcomm}>." return fi # avoid selecting a recovered node as bootstrap if possible for node in $all_nodes; do local pcmk_node=$(galera_to_pcmk_name $node) if [ -z "$pcmk_node" ]; then ocf_log err "Could not determine pacemaker node from galera name <${node}>." return else node=$pcmk_node fi if is_no_grastate $node; then nodes_recovered="$nodes_recovered $node" else nodes="$nodes $node" fi done for node in $nodes_recovered $nodes; do # On clean shutdown, galera sets the last stopped node as 'safe to bootstrap', # so use this hint when we can safe_to_bootstrap=$(get_safe_to_bootstrap $node) # Special case for 2-node clusters: during a network split, rely on # pacemaker's quorum to check whether we can restart galera if [ "$safe_to_bootstrap" != "1" ] && [ "$node" = "$NODENAME" ] && is_two_node_mode_active; then is_last_node_in_quorate_partition if [ $? -eq 0 ]; then ocf_log warn "Survived a split in a 2-node cluster, considering ourselves safe to bootstrap" safe_to_bootstrap=1 fi fi if [ "$safe_to_bootstrap" = "1" ]; then # Galera marked the node as safe to boostrap during shutdown. Let's just # pick it as our bootstrap node. ocf_log info "Node <${node}> is marked as safe to bootstrap." best_node=$node # We don't need to wait for the other nodes to report state in this case missing_nodes=0 break fi last_commit=$(get_last_commit $node) if [ -z "$last_commit" ]; then ocf_log info "Waiting on node <${node}> to report database status before Master instances can start." missing_nodes=1 continue fi # this means -1, or that no commit has occured yet. if [ "$last_commit" = "18446744073709551615" ]; then last_commit="0" fi greater_than_equal_long "$last_commit" "$best_commit" if [ $? -eq 0 ]; then best_node=$(ocf_attribute_target $node) best_commit=$last_commit fi done if [ $missing_nodes -eq 1 ]; then return fi ocf_log info "Promoting $best_node to be our bootstrap node" set_bootstrap_node $best_node set_master_score $best_node } detect_safe_to_bootstrap() { local safe_to_bootstrap="" local uuid="" local seqno="" if [ -f ${OCF_RESKEY_datadir}/grastate.dat ]; then ocf_log info "attempting to read safe_to_bootstrap flag from ${OCF_RESKEY_datadir}/grastate.dat" safe_to_bootstrap=$(sed -n 's/^safe_to_bootstrap:\s*\(.*\)$/\1/p' < ${OCF_RESKEY_datadir}/grastate.dat) uuid=$(sed -n 's/^uuid:\s*\(.*\)$/\1/p' < ${OCF_RESKEY_datadir}/grastate.dat) seqno=$(sed -n 's/^seqno:\s*\(.*\)$/\1/p' < ${OCF_RESKEY_datadir}/grastate.dat) fi if [ -z "$uuid" ] || \ [ "$uuid" = "00000000-0000-0000-0000-000000000000" ]; then clear_safe_to_bootstrap return fi if [ "$safe_to_bootstrap" = "1" ]; then if [ -z "$seqno" ] || [ "$seqno" = "-1" ]; then clear_safe_to_bootstrap return fi fi if [ "$safe_to_bootstrap" = "1" ] || [ "$safe_to_bootstrap" = "0" ]; then set_safe_to_bootstrap $safe_to_bootstrap else clear_safe_to_bootstrap fi } detect_last_commit() { local last_commit local recover_args="--defaults-file=$OCF_RESKEY_config \ --pid-file=$OCF_RESKEY_pid \ --socket=$OCF_RESKEY_socket \ --datadir=$OCF_RESKEY_datadir" local recovery_file_regex='s/.*WSREP\:.*position\s*recovery.*--log_error='\''\([^'\'']*\)'\''.*/\1/p' local recovered_position_regex='s/.*WSREP\:\s*[R|r]ecovered\s*position.*\:\(.*\)\s*$/\1/p' # codership/galera#354 # Some ungraceful shutdowns can leave an empty gvwstate.dat on # disk. This will prevent galera to join the cluster if it is # configured to attempt PC recovery. Removing that file makes the # node fall back to the normal, unoptimized joining process. if [ -f ${OCF_RESKEY_datadir}/gvwstate.dat ] && \ [ ! -s ${OCF_RESKEY_datadir}/gvwstate.dat ]; then ocf_log warn "empty ${OCF_RESKEY_datadir}/gvwstate.dat detected, removing it to prevent PC recovery failure at next restart" rm -f ${OCF_RESKEY_datadir}/gvwstate.dat fi ocf_log info "attempting to detect last commit version by reading ${OCF_RESKEY_datadir}/grastate.dat" last_commit="$(cat ${OCF_RESKEY_datadir}/grastate.dat | sed -n 's/^seqno.\s*\(.*\)\s*$/\1/p')" if [ -z "$last_commit" ] || [ "$last_commit" = "-1" ]; then local tmp=$(mktemp) chown $OCF_RESKEY_user:$OCF_RESKEY_group $tmp # if we pass here because grastate.dat doesn't exist, # try not to bootstrap from this node if possible if [ ! -f ${OCF_RESKEY_datadir}/grastate.dat ]; then set_no_grastate fi ocf_log info "now attempting to detect last commit version using 'mysqld_safe --wsrep-recover'" $SU - $OCF_RESKEY_user -s /bin/sh -c \ "${OCF_RESKEY_binary} $recover_args --wsrep-recover --log-error=$tmp 2>/dev/null" last_commit="$(cat $tmp | sed -n $recovered_position_regex | tail -1)" if [ -z "$last_commit" ]; then # Galera uses InnoDB's 2pc transactions internally. If # server was stopped in the middle of a replication, the # recovery may find a "prepared" XA transaction in the # redo log, and mysql won't recover automatically local recovery_file="$(cat $tmp | sed -n $recovery_file_regex)" if [ -e $recovery_file ]; then cat $recovery_file | grep -q -E '\[ERROR\]\s+Found\s+[0-9]+\s+prepared\s+transactions!' 2>/dev/null if [ $? -eq 0 ]; then # we can only rollback the transaction, but that's OK # since the DB will get resynchronized anyway ocf_log warn "local node <${NODENAME}> was not shutdown properly. Rollback stuck transaction with --tc-heuristic-recover" $SU - $OCF_RESKEY_user -s /bin/sh -c \ "${OCF_RESKEY_binary} $recover_args --wsrep-recover \ --tc-heuristic-recover=rollback --log-error=$tmp 2>/dev/null" last_commit="$(cat $tmp | sed -n $recovered_position_regex | tail -1)" if [ ! -z "$last_commit" ]; then ocf_log warn "State recovered. force SST at next restart for full resynchronization" rm -f ${OCF_RESKEY_datadir}/grastate.dat # try not to bootstrap from this node if possible set_no_grastate fi fi fi fi rm -f $tmp fi if [ ! -z "$last_commit" ]; then ocf_log info "Last commit version found: $last_commit" set_last_commit $last_commit return $OCF_SUCCESS else ocf_exit_reason "Unable to detect last known write sequence number" clear_last_commit return $OCF_ERR_GENERIC fi } # For galera, promote is really start galera_promote() { local rc local extra_opts local bootstrap local safe_to_bootstrap master_exists if [ $? -eq 0 ]; then # join without bootstrapping extra_opts="--wsrep-cluster-address=${OCF_RESKEY_wsrep_cluster_address}" else bootstrap=$(is_bootstrap) if ocf_is_true $bootstrap; then # The best node for bootstrapping wasn't cleanly shutdown. Allow # bootstrapping anyways if [ "$(get_safe_to_bootstrap)" = "0" ]; then sed -ie 's/^\(safe_to_bootstrap:\) 0/\1 1/' ${OCF_RESKEY_datadir}/grastate.dat ocf_log info "safe_to_bootstrap in ${OCF_RESKEY_datadir}/grastate.dat set to 1 on node ${NODENAME}" fi ocf_log info "Node <${NODENAME}> is bootstrapping the cluster" extra_opts="--wsrep-cluster-address=gcomm://" else # We are being promoted without having the bootstrap # attribute in the CIB, which means we are supposed to # join a cluster; however if we end up here, there is no # Master remaining right now, which means there is no # cluster to join anymore. So force a demotion, and and # let the RA decide later which node should be the next # bootstrap node. ocf_log warn "There is no running cluster to join, demoting ourself" clear_master_score return $OCF_SUCCESS fi fi galera_monitor if [ $? -eq $OCF_RUNNING_MASTER ]; then if ocf_is_true $bootstrap; then promote_everyone clear_bootstrap_node ocf_log info "boostrap node already up, promoting the rest of the galera instances." fi clear_safe_to_bootstrap clear_last_commit return $OCF_SUCCESS fi # last commit/safe_to_bootstrap flag are no longer relevant once promoted clear_last_commit clear_safe_to_bootstrap mysql_common_prepare_dirs mysql_common_start "$extra_opts" rc=$? if [ $rc != $OCF_SUCCESS ]; then return $rc fi # At this point, the mysql pidfile is created on disk and the # mysql server is reacheable via its UNIX socket. If we are a # joiner, SST transfers (rsync) have finished, but an IST may # still be requested or ongoing galera_monitor rc=$? if [ $rc != $OCF_SUCCESS -a $rc != $OCF_RUNNING_MASTER ]; then ocf_exit_reason "Failed initial monitor action" return $rc fi is_readonly if [ $? -eq 0 ]; then ocf_exit_reason "Failure. Master instance started in read-only mode, check configuration." return $OCF_ERR_GENERIC fi if ocf_is_true $bootstrap; then promote_everyone clear_bootstrap_node # clear attribute no-grastate. if last shutdown was # not clean, we cannot be extra-cautious by requesting a SST # since this is the bootstrap node clear_no_grastate ocf_log info "Bootstrap complete, promoting the rest of the galera instances." else # if this is not the bootstrap node, make sure this instance # syncs with the rest of the cluster before promotion returns. wait_for_sync # sync is done, clear info about last startup clear_no_grastate fi ocf_log info "Galera started" return $OCF_SUCCESS } galera_demote() { mysql_common_stop rc=$? if [ $rc -ne $OCF_SUCCESS ] && [ $rc -ne $OCF_NOT_RUNNING ]; then ocf_exit_reason "Failed to stop Master galera instance during demotion to Master" return $rc fi # if this node was previously a bootstrap node, that is no longer the case. clear_bootstrap_node clear_last_commit clear_no_grastate clear_safe_to_bootstrap # Clear master score here rather than letting pacemaker do so once # demote finishes. This way a promote cannot take place right # after this demote even if pacemaker is requested to do so. It # will first have to run a start/monitor op, to reprobe the state # of the other galera nodes and act accordingly. clear_master_score # record last commit for next promotion detect_safe_to_bootstrap detect_last_commit rc=$? return $rc } galera_start() { local rc local galera_node galera_node=$(pcmk_to_galera_name $NODENAME) if [ -z "$galera_node" ]; then ocf_exit_reason "Could not determine galera name from pacemaker node <${NODENAME}>." return $OCF_ERR_CONFIGURED fi echo $OCF_RESKEY_wsrep_cluster_address | grep -q -F $galera_node if [ $? -ne 0 ]; then ocf_exit_reason "local node <${NODENAME}> (galera node <${galera_node}>) must be a member of the wsrep_cluster_address <${OCF_RESKEY_wsrep_cluster_address}> to start this galera instance" return $OCF_ERR_CONFIGURED fi galera_monitor if [ $? -eq $OCF_RUNNING_MASTER ]; then ocf_exit_reason "master galera instance started outside of the cluster's control" return $OCF_ERR_GENERIC fi mysql_common_prepare_dirs detect_safe_to_bootstrap detect_last_commit rc=$? if [ $rc -ne $OCF_SUCCESS ]; then return $rc fi master_exists if [ $? -eq 0 ]; then ocf_log info "Master instances are already up, setting master score so this instance will join galera cluster." set_master_score $NODENAME else clear_master_score detect_first_master fi return $OCF_SUCCESS } galera_monitor() { local rc local galera_node local status_loglevel="err" # Set loglevel to info during probe if ocf_is_probe; then status_loglevel="info" fi mysql_common_status $status_loglevel rc=$? if [ $rc -eq $OCF_NOT_RUNNING ]; then last_commit=$(get_last_commit $node) if [ -n "$last_commit" ]; then # if last commit is set, this instance is considered started in slave mode rc=$OCF_SUCCESS master_exists if [ $? -ne 0 ]; then detect_first_master else # a master instance exists and is healthy, promote this # local read only instance # so it can join the master galera cluster. set_master_score fi fi return $rc elif [ $rc -ne $OCF_SUCCESS ]; then return $rc fi # if we make it here, mysql is running. Check cluster status now. galera_node=$(pcmk_to_galera_name $NODENAME) if [ -z "$galera_node" ]; then ocf_exit_reason "Could not determine galera name from pacemaker node <${NODENAME}>." return $OCF_ERR_CONFIGURED fi echo $OCF_RESKEY_wsrep_cluster_address | grep -q -F $galera_node if [ $? -ne 0 ]; then ocf_exit_reason "local node <${NODENAME}> (galera node <${galera_node}>) is started, but is not a member of the wsrep_cluster_address <${OCF_RESKEY_wsrep_cluster_address}>" return $OCF_ERR_GENERIC fi is_primary if [ $? -eq 0 ]; then if ocf_is_probe; then # restore master score during probe # if we detect this is a master instance set_master_score fi rc=$OCF_RUNNING_MASTER else # It seems that with recent galera (26.4+), a joiner that is # connected to a Primary component and is preparing its IST # request might still temporarily report its state as # Non-Primary. Do not fail in this case as the promote # operation will loop until the IST finishes or the promote # times out. if [ "$__OCF_ACTION" = "promote" ] && ! ocf_is_true $(is_bootstrap); then ocf_log info "local node <${NODENAME}> is receiving a State Transfer." else ocf_exit_reason "local node <${NODENAME}> is started, but not in primary mode. Unknown state." rc=$OCF_ERR_GENERIC fi fi return $rc } galera_stop() { local rc # make sure the process is stopped mysql_common_stop rc=$1 clear_safe_to_bootstrap clear_last_commit clear_master_score clear_bootstrap_node clear_no_grastate return $rc } galera_validate() { if [ "$OCF_CHECK_LEVEL" -eq 10 ]; then if ! ocf_is_ms; then ocf_exit_reason "Galera must be configured as a multistate Master/Slave resource." return $OCF_ERR_CONFIGURED fi fi if [ -z "$OCF_RESKEY_wsrep_cluster_address" ]; then ocf_exit_reason "Galera must be configured with a wsrep_cluster_address value." return $OCF_ERR_CONFIGURED fi mysql_common_validate } case "$1" in meta-data) meta_data exit $OCF_SUCCESS;; usage|help) usage exit $OCF_SUCCESS;; esac [ "$__OCF_ACTION" = "start" ] && OCF_CHECK_LEVEL=10 galera_validate rc=$? LSB_STATUS_STOPPED=3 if [ $rc -ne 0 ]; then case "$1" in stop) exit $OCF_SUCCESS;; monitor) exit $OCF_NOT_RUNNING;; status) exit $LSB_STATUS_STOPPED;; *) exit $rc;; esac fi if [ -z "${OCF_RESKEY_check_passwd}" ]; then # This value is automatically sourced from /etc/sysconfig/checkcluster if available OCF_RESKEY_check_passwd=${MYSQL_PASSWORD} fi if [ -z "${OCF_RESKEY_check_user}" ]; then # This value is automatically sourced from /etc/sysconfig/checkcluster if available OCF_RESKEY_check_user=${MYSQL_USERNAME} fi : ${OCF_RESKEY_check_user="root"} MYSQL_OPTIONS_CHECK="-nNE --user=${OCF_RESKEY_check_user}" if ocf_is_true "${OCF_RESKEY_check_passwd_use_empty}"; then MYSQL_OPTIONS_CHECK="$MYSQL_OPTIONS_CHECK --password=" elif [ -n "${OCF_RESKEY_check_passwd}" ]; then MYSQL_OPTIONS_CHECK="$MYSQL_OPTIONS_CHECK --password=${OCF_RESKEY_check_passwd}" fi # This value is automatically sourced from /etc/sysconfig/checkcluster if available if [ -n "${MYSQL_HOST}" ]; then MYSQL_OPTIONS_CHECK="$MYSQL_OPTIONS_CHECK -h ${MYSQL_HOST}" fi # This value is automatically sourced from /etc/sysconfig/checkcluster if available if [ -n "${MYSQL_PORT}" ]; then MYSQL_OPTIONS_CHECK="$MYSQL_OPTIONS_CHECK -P ${MYSQL_PORT}" fi +ocf_version_cmp "$OCF_RESKEY_crm_feature_set" "3.1.0" +res=$? +if [ -z "$OCF_RESKEY_crm_feature_set" ] || [ $res -eq 2 ]; then + XMLOPT="--output-as=xml" + ocf_version_cmp "$OCF_RESKEY_crm_feature_set" "3.2.0" + if [ $? -eq 1 ]; then + crm_mon_no_validation -1 $XMLOPT >/dev/null 2>&1 + if [ $? -ne 0 ]; then + XMLOPT="--as-xml" + fi + fi +else + XMLOPT="--as-xml" +fi # What kind of method was invoked? case "$1" in start) galera_start;; stop) galera_stop;; status) mysql_common_status err;; monitor) galera_monitor;; promote) galera_promote;; demote) galera_demote;; validate-all) exit $OCF_SUCCESS;; *) usage exit $OCF_ERR_UNIMPLEMENTED;; esac # vi:sw=4:ts=4:et: diff --git a/heartbeat/ocf-shellfuncs.in b/heartbeat/ocf-shellfuncs.in index 96f612095..9335cbf00 100644 --- a/heartbeat/ocf-shellfuncs.in +++ b/heartbeat/ocf-shellfuncs.in @@ -1,1104 +1,1110 @@ # # # Common helper functions for the OCF Resource Agents supplied by # heartbeat. # # Copyright (c) 2004 SUSE LINUX AG, Lars Marowsky-Brée # All Rights Reserved. # # # This library is free software; you can redistribute it and/or # modify it under the terms of the GNU Lesser General Public # License as published by the Free Software Foundation; either # version 2.1 of the License, or (at your option) any later version. # # This library is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU # Lesser General Public License for more details. # # You should have received a copy of the GNU Lesser General Public # License along with this library; if not, write to the Free Software # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA # # Build version: $Format:%H$ # TODO: Some of this should probably split out into a generic OCF # library for shell scripts, but for the time being, we'll just use it # ourselves... # # TODO wish-list: # - Generic function for evaluating version numbers # - Generic function(s) to extract stuff from our own meta-data # - Logging function which automatically adds resource identifier etc # prefixes # TODO: Move more common functionality for OCF RAs here. # # This was common throughout all legacy Heartbeat agents unset LC_ALL; export LC_ALL unset LANGUAGE; export LANGUAGE : ${HA_SBIN_DIR:=@sbindir@} __SCRIPT_NAME=`basename $0` if [ -z "$OCF_ROOT" ]; then : ${OCF_ROOT=@OCF_ROOT_DIR@} fi if [ "$OCF_FUNCTIONS_DIR" = ${OCF_ROOT}/resource.d/heartbeat ]; then # old unset OCF_FUNCTIONS_DIR fi : ${OCF_FUNCTIONS_DIR:=${OCF_ROOT}/lib/heartbeat} . ${OCF_FUNCTIONS_DIR}/ocf-binaries . ${OCF_FUNCTIONS_DIR}/ocf-returncodes . ${OCF_FUNCTIONS_DIR}/ocf-directories . ${OCF_FUNCTIONS_DIR}/ocf-rarun . ${OCF_FUNCTIONS_DIR}/ocf-distro # Define OCF_RESKEY_CRM_meta_interval in case it isn't already set, # to make sure that ocf_is_probe() always works : ${OCF_RESKEY_CRM_meta_interval=0} ocf_is_root() { if [ X`id -u` = X0 ]; then true else false fi } ocf_maybe_random() { if test -c /dev/urandom; then od -An -N4 -tu4 /dev/urandom | tr -d '[:space:]' else awk -v pid=$$ 'BEGIN{srand(pid); print rand()}' | sed 's/^.*[.]//' fi } # Portability comments: # o The following rely on Bourne "sh" pattern-matching, which is usually # that for filename generation (note: not regexp). # o The "*) true ;;" clause is probably unnecessary, but is included # here for completeness. # o The negation in the pattern uses "!". This seems to be common # across many OSes (whereas the alternative "^" fails on some). # o If an OS is encountered where this negation fails, then a possible # alternative would be to replace the function contents by (e.g.): # [ -z "`echo $1 | tr -d '[0-9]'`" ] # ocf_is_decimal() { case "$1" in ""|*[!0-9]*) # empty, or at least one non-decimal false ;; *) true ;; esac } ocf_is_true() { case "$1" in yes|true|1|YES|TRUE|True|ja|on|ON) true ;; *) false ;; esac } ocf_is_hex() { case "$1" in ""|*[!0-9a-fA-F]*) # empty, or at least one non-hex false ;; *) true ;; esac } ocf_is_octal() { case "$1" in ""|*[!0-7]*) # empty, or at least one non-octal false ;; *) true ;; esac } __ocf_set_defaults() { __OCF_ACTION="$1" # Return to sanity for the agents... unset LANG LC_ALL=C export LC_ALL # TODO: Review whether we really should source this. Or rewrite # to match some emerging helper function syntax...? This imports # things which no OCF RA should be using... # Strip the OCF_RESKEY_ prefix from this particular parameter if [ -z "$OCF_RESKEY_OCF_CHECK_LEVEL" ]; then : ${OCF_CHECK_LEVEL:=0} else : ${OCF_CHECK_LEVEL:=$OCF_RESKEY_OCF_CHECK_LEVEL} fi if [ ! -d "$OCF_ROOT" ]; then ha_log "ERROR: OCF_ROOT points to non-directory $OCF_ROOT." exit $OCF_ERR_GENERIC fi if [ -z "$OCF_RESOURCE_TYPE" ]; then : ${OCF_RESOURCE_TYPE:=$__SCRIPT_NAME} fi if [ "x$__OCF_ACTION" = "xmeta-data" ]; then : ${OCF_RESOURCE_INSTANCE:="RESOURCE_ID"} fi if [ -z "$OCF_RA_VERSION_MAJOR" ]; then : We are being invoked as an init script. : Fill in some things with reasonable values. : ${OCF_RESOURCE_INSTANCE:="default"} return 0 fi if [ -z "$OCF_RESOURCE_INSTANCE" ]; then ha_log "ERROR: Need to tell us our resource instance name." exit $OCF_ERR_ARGS fi } hadate() { date "+${HA_DATEFMT}" } set_logtag() { if [ -z "$HA_LOGTAG" ]; then if [ -n "$OCF_RESOURCE_INSTANCE" ]; then HA_LOGTAG="$__SCRIPT_NAME($OCF_RESOURCE_INSTANCE)[$$]" else HA_LOGTAG="$__SCRIPT_NAME[$$]" fi fi } __ha_log() { local ignore_stderr=false local loglevel [ "x$1" = "x--ignore-stderr" ] && ignore_stderr=true && shift [ none = "$HA_LOGFACILITY" ] && HA_LOGFACILITY="" # if we're connected to a tty, then output to stderr if tty >/dev/null; then if [ "x$HA_debug" = "x0" -a "x$loglevel" = xdebug ] ; then return 0 elif [ "$ignore_stderr" = "true" ]; then # something already printed this error to stderr, so ignore return 0 fi if [ "$HA_LOGTAG" ]; then echo "$HA_LOGTAG: $*" else echo "$*" fi >&2 return 0 fi set_logtag if [ "x${HA_LOGD}" = "xyes" ] ; then ha_logger -t "${HA_LOGTAG}" "$@" if [ "$?" -eq "0" ] ; then return 0 fi fi if [ -n "$HA_LOGFACILITY" ] then : logging through syslog # loglevel is unknown, use 'notice' for now loglevel=notice case "${*}" in *ERROR*) loglevel=err;; *WARN*) loglevel=warning;; *INFO*|info) loglevel=info;; esac logger -t "$HA_LOGTAG" -p ${HA_LOGFACILITY}.${loglevel} "${*}" fi if [ -n "$HA_LOGFILE" ] then : appending to $HA_LOGFILE echo `hadate`" $HA_LOGTAG: ${*}" >> $HA_LOGFILE fi if [ -z "$HA_LOGFACILITY" -a -z "$HA_LOGFILE" ] && ! [ "$ignore_stderr" = "true" ] then : appending to stderr echo `hadate`"${*}" >&2 fi if [ -n "$HA_DEBUGLOG" ] then : appending to $HA_DEBUGLOG if [ "$HA_LOGFILE"x != "$HA_DEBUGLOG"x ]; then echo "$HA_LOGTAG: "`hadate`"${*}" >> $HA_DEBUGLOG fi fi } ha_log() { __ha_log "$@" } ha_debug() { if [ "x${HA_debug}" = "x0" ] || [ -z "${HA_debug}" ] ; then return 0 fi if tty >/dev/null; then if [ "$HA_LOGTAG" ]; then echo "$HA_LOGTAG: $*" else echo "$*" fi >&2 return 0 fi set_logtag if [ "x${HA_LOGD}" = "xyes" ] ; then ha_logger -t "${HA_LOGTAG}" -D "ha-debug" "$@" if [ "$?" -eq "0" ] ; then return 0 fi fi [ none = "$HA_LOGFACILITY" ] && HA_LOGFACILITY="" if [ -n "$HA_LOGFACILITY" ] then : logging through syslog logger -t "$HA_LOGTAG" -p "${HA_LOGFACILITY}.debug" "${*}" fi if [ -n "$HA_DEBUGLOG" ] then : appending to $HA_DEBUGLOG echo "$HA_LOGTAG: "`hadate`"${*}" >> $HA_DEBUGLOG fi if [ -z "$HA_LOGFACILITY" -a -z "$HA_DEBUGLOG" ] then : appending to stderr echo "$HA_LOGTAG: `hadate`${*}: ${HA_LOGFACILITY}" >&2 fi } ha_parameter() { local VALUE VALUE=`sed -e 's%[ ][ ]*% %' -e 's%^ %%' -e 's%#.*%%' $HA_CF | grep -i "^$1 " | sed 's%[^ ]* %%'` if [ "X$VALUE" = X ] then case $1 in keepalive) VALUE=2;; deadtime) ka=`ha_parameter keepalive` VALUE=`expr $ka '*' 2 '+' 1`;; esac fi echo $VALUE } ocf_log() { # TODO: Revisit and implement internally. if [ $# -lt 2 ] then ocf_log err "Not enough arguments [$#] to ocf_log." fi __OCF_PRIO="$1" shift __OCF_MSG="$*" case "${__OCF_PRIO}" in crit) __OCF_PRIO="CRIT";; err) __OCF_PRIO="ERROR";; warn) __OCF_PRIO="WARNING";; info) __OCF_PRIO="INFO";; debug)__OCF_PRIO="DEBUG";; *) __OCF_PRIO=`echo ${__OCF_PRIO}| tr '[a-z]' '[A-Z]'`;; esac if [ "${__OCF_PRIO}" = "DEBUG" ]; then ha_debug "${__OCF_PRIO}: $__OCF_MSG" else ha_log "${__OCF_PRIO}: $__OCF_MSG" fi } # # ocf_exit_reason: print exit error string to stderr # Usage: Allows the OCF script to provide a string # describing why the exit code was returned. # Arguments: reason - required, The string that represents why the error # occured. # ocf_exit_reason() { local cookie="$OCF_EXIT_REASON_PREFIX" local fmt local msg # No argument is likely not intentional. # Just one argument implies a printf format string of just "%s". # "Least surprise" in case some interpolated string from variable # expansion or other contains a percent sign. # More than one argument: first argument is going to be the format string. case $# in 0) ocf_log err "Not enough arguments to ocf_log_exit_msg." ;; 1) fmt="%s" ;; *) fmt=$1 shift case $fmt in *%*) : ;; # ok, does look like a format string *) ocf_log warn "Does not look like format string: [$fmt]" ;; esac ;; esac if [ -z "$cookie" ]; then # use a default prefix cookie="ocf-exit-reason:" fi msg=$(printf "${fmt}" "$@") printf >&2 "%s%s\n" "$cookie" "$msg" __ha_log --ignore-stderr "ERROR: $msg" } # # ocf_deprecated: Log a deprecation warning # Usage: ocf_deprecated [param-name] # Arguments: param-name optional, name of a boolean resource # parameter that can be used to suppress # the warning (default # "ignore_deprecation") ocf_deprecated() { local param param=${1:-ignore_deprecation} # don't use ${!param} here, it's a bashism if ! ocf_is_true $(eval echo \$OCF_RESKEY_$param); then ocf_log warn "This resource agent is deprecated" \ "and may be removed in a future release." \ "See the man page for details." \ "To suppress this warning, set the \"${param}\"" \ "resource parameter to true." fi } # # Ocf_run: Run a script, and log its output. # Usage: ocf_run [-q] [-info|-warn|-err] # -q: don't log the output of the command if it succeeds # -info|-warn|-err: log the output of the command at given # severity if it fails (defaults to err) # ocf_run() { local rc local output local verbose=1 local loglevel=err local var for var in 1 2 do case "$1" in "-q") verbose="" shift 1;; "-info"|"-warn"|"-err"|"-debug") loglevel=${1#-} shift 1;; *) ;; esac done output=`"$@" 2>&1` rc=$? [ -n "$output" ] && output="$(echo "$output" | tr -s ' \t\r\n' ' ')" if [ $rc -eq 0 ]; then if [ "$verbose" -a ! -z "$output" ]; then ocf_log info "$output" fi else if [ ! -z "$output" ]; then ocf_log $loglevel "$output" else ocf_log $loglevel "command failed: $*" fi fi return $rc } ocf_pidfile_status() { local pid pidfile="$1" if [ ! -e "$pidfile" ]; then # Not exists return 2 fi pid=$(cat "$pidfile") kill -0 "$pid" > /dev/null 2>&1 if [ $? = 0 ]; then return 0 fi # Stale return 1 } # mkdir(1) based locking # first the directory is created with the name given as $1 # then a file named "pid" is created within that directory with # the process PID # stale locks are handled carefully, the inode of a directory # needs to match before and after test if the process is running # empty directories are also handled appropriately # we relax (sleep) occasionally to allow for other processes to # finish managing the lock in case they are in the middle of the # business relax() { sleep 0.5; } ocf_get_stale_pid() { local piddir pid dir_inode piddir="$1" [ -z "$piddir" ] && return 2 dir_inode="`ls -di $piddir 2>/dev/null`" [ -z "$dir_inode" ] && return 1 pid=`cat $piddir/pid 2>/dev/null` if [ -z "$pid" ]; then # empty directory? relax if [ "$dir_inode" = "`ls -di $piddir 2>/dev/null`" ]; then echo $dir_inode else return 1 fi elif kill -0 $pid >/dev/null 2>&1; then return 1 elif relax && [ -e "$piddir/pid" ] && [ "$dir_inode" = "`ls -di $piddir 2>/dev/null`" ]; then echo $pid else return 1 fi } # There is a race when the following two functions to manage the # lock file (mk and rm) are invoked in parallel by different # instances. It is up to the caller to reduce probability of that # taking place (see ocf_take_lock() below). ocf_mk_pid() { mkdir $1 2>/dev/null && echo $$ > $1/pid } ocf_rm_pid() { rm -f $1/pid rmdir $1 2>/dev/null } # Testing and subsequently removing a stale lock (containing the # process pid) is inherently difficult to do in such a way as to # prevent a race between creating a pid file and removing it and # its directory. We reduce the probability of that happening by # checking if the stale lock persists over a random period of # time. ocf_take_lock() { local lockdir=$1 local rnd local stale_pid # we don't want it too short, so strip leading zeros rnd=$(ocf_maybe_random | sed 's/^0*//') stale_pid=`ocf_get_stale_pid $lockdir` if [ -n "$stale_pid" ]; then sleep 0.$rnd # remove "stale pid" only if it persists [ "$stale_pid" = "`ocf_get_stale_pid $lockdir`" ] && ocf_rm_pid $lockdir fi while ! ocf_mk_pid $lockdir; do ocf_log info "Sleeping until $lockdir is released..." sleep 0.$rnd done } ocf_release_lock_on_exit() { trap "ocf_rm_pid $1" EXIT } # returns true if the CRM is currently running a probe. A probe is # defined as a monitor operation with a monitoring interval of zero. ocf_is_probe() { [ "$__OCF_ACTION" = "monitor" -a "$OCF_RESKEY_CRM_meta_interval" = 0 ] } # returns true if the resource is configured as a clone. This is # defined as a resource where the clone-max meta attribute is present. ocf_is_clone() { [ ! -z "${OCF_RESKEY_CRM_meta_clone_max}" ] } # returns true if the resource is configured as a multistate # (master/slave) resource. This is defined as a resource where the # master-max meta attribute is present, and set to greater than zero. ocf_is_ms() { [ "${OCF_RESKEY_CRM_meta_promotable}" = "true" ] || { [ ! -z "${OCF_RESKEY_CRM_meta_master_max}" ] && [ "${OCF_RESKEY_CRM_meta_master_max}" -gt 0 ]; } } # version check functions # allow . and - to delimit version numbers # max version number is 999 # ocf_is_ver() { echo $1 | grep '^[0-9][0-9.-]*[0-9A-Za-z.\+-]*$' >/dev/null 2>&1 } # usage: ocf_version_cmp VER1 VER2 # version strings can contain digits, dots, and dashes # must start and end with a digit # returns: # 0: VER1 smaller (older) than VER2 # 1: versions equal # 2: VER1 greater (newer) than VER2 # 3: bad format ocf_version_cmp() { ocf_is_ver "$1" || return 3 ocf_is_ver "$2" || return 3 local v1=$1 local v2=$2 sort_version="sort -t. -k 1,1n -k 2,2n -k 3,3n -k 4,4n" older=$( (echo "$v1"; echo "$v2") | $sort_version | head -1 ) if [ "$v1" = "$v2" ]; then return 1 elif [ "$v1" = "$older" ]; then return 0 else return 2 # -1 would look funny in shell ;-) fi } ocf_local_nodename() { # use crm_node -n for pacemaker > 1.1.8 which pacemakerd > /dev/null 2>&1 if [ $? -eq 0 ]; then local version=$(pacemakerd -$ | awk '/^Pacemaker /{ print $2 }') version=$(echo $version | awk -F- '{ print $1 }') ocf_version_cmp "$version" "1.1.8" if [ $? -eq 2 ]; then which crm_node > /dev/null 2>&1 if [ $? -eq 0 ]; then crm_node -n return fi fi fi # otherwise use uname -n uname -n } # usage: dirname DIR dirname() { local a local b [ $# = 1 ] || return 1 a="$1" while [ 1 ]; do b="${a%/}" [ "$a" = "$b" ] && break a="$b" done b=${a%/*} [ -z "$b" -o "$a" = "$b" ] && b="." echo "$b" return 0 } # usage: systemd_is_running # returns: # 0 PID 1 is systemd # 1 otherwise systemd_is_running() { [ "$(cat /proc/1/comm 2>/dev/null)" = "systemd" ] } # usage: systemd_drop_in systemd_drop_in() { local conf_file if [ $# -ne 3 ]; then ocf_log err "Incorrect number of arguments [$#] for systemd_drop_in." fi systemdrundir="/run/systemd/system/resource-agents-deps.target.d" mkdir -p "$systemdrundir" conf_file="$systemdrundir/$1.conf" cat >"$conf_file" <&1) rc=$? ocf_log debug "result: $result" [ $rc -eq 0 ] && break sleep $sleep done if [ $rc -ne 0 ]; then ocf_exit_reason "curl $args failed $tries tries" exit $OCF_ERR_GENERIC fi echo "$result" return $rc } # move process to root cgroup if realtime scheduling is enabled ocf_move_to_root_cgroup_if_rt_enabled() { if [ -e "/sys/fs/cgroup/cpu/cpu.rt_runtime_us" ]; then echo $$ >> /sys/fs/cgroup/cpu/tasks if [ "$?" -ne "0" ]; then ocf_log warn "Unable to move PID $$ to the root cgroup" fi fi } # usage: crm_mon_no_validation args... # run crm_mon without any cib schema validation # This is useful when an agent runs in a bundle to avoid potential # schema validation errors when host and bundle are not perfectly aligned # To be used, your shell must support on process substitution (e.g. bash) # returns: # crm_mon_no_validation() { # The subshell prevents parsing error with incompatible shells - "$SHELL" -c "CIB_file=<(${HA_SBIN_DIR}/cibadmin -Q | sed 's/validate-with=\"[^\"]*\"/validate-with=\"none\"/') \ - ${HA_SBIN_DIR}/crm_mon \$*" -- $* + ocf_version_cmp "$OCF_RESKEY_crm_feature_set" "3.19.7" + if [ $res -eq 2 ] || [ $res -eq 1 ]; then + "$SHELL" -c "CIB_file=<(${HA_SBIN_DIR}/cibadmin -Q \ + ${HA_SBIN_DIR}/crm_mon \$*" -- $* + else + "$SHELL" -c "CIB_file=<(${HA_SBIN_DIR}/cibadmin -Q | sed 's/validate-with=\"[^\"]*\"/validate-with=\"none\"/') \ + ${HA_SBIN_DIR}/crm_mon \$*" -- $* + fi } # # pseudo_resource status tracking function... # # This allows pseudo resources to give correct status information. As we add # resource monitoring, and better resource tracking in general, this will # become essential. # # These scripts work because ${HA_RSCTMP} is cleaned on node reboot. # # We create "resource-string" tracking files under ${HA_RSCTMP} in a # very simple way: # # Existence of "${HA_RSCTMP}/resource-string" means that we consider # the resource named by "resource-string" to be running. # # Note that "resource-string" needs to be unique. Using the resource type # plus the resource instance arguments to make up the resource string # is probably sufficient... # # usage: ha_pseudo_resource resource-string op [tracking_file] # where op is {start|stop|monitor|status|restart|reload|print} # print is a special op which just prints the tracking file location # user can override our choice of the tracking file location by # specifying it as the third arg # Note that all operations are silent... # ha_pseudo_resource() { local ha_resource_tracking_file="${3:-${HA_RSCTMP}/$1}" case $2 in start|restart|reload) touch "$ha_resource_tracking_file";; stop) rm -f "$ha_resource_tracking_file";; status|monitor) if [ -f "$ha_resource_tracking_file" ] then return 0 else case $2 in status) return 3;; *) return 7;; esac fi;; print) echo "$ha_resource_tracking_file";; *) return 3;; esac } # usage: rmtempdir TMPDIR rmtempdir() { [ $# = 1 ] || return 1 if [ -e "$1" ]; then rmdir "$1" || return 1 fi return 0 } # usage: maketempfile [-d] maketempfile() { if [ $# = 1 -a "$1" = "-d" ]; then mktemp -d return 0 elif [ $# != 0 ]; then return 1 fi mktemp return 0 } # usage: rmtempfile TMPFILE rmtempfile () { [ $# = 1 ] || return 1 if [ -e "$1" ]; then rm "$1" || return 1 fi return 0 } # echo the first lower supported check level # pass set of levels supported by the agent # (in increasing order, 0 is optional) ocf_check_level() { local lvl prev lvl=0 prev=0 if ocf_is_decimal "$OCF_CHECK_LEVEL"; then # the level list should be very short for lvl; do if [ "$lvl" -eq "$OCF_CHECK_LEVEL" ]; then break elif [ "$lvl" -gt "$OCF_CHECK_LEVEL" ]; then lvl=$prev # the previous one break fi prev=$lvl done fi echo $lvl } # usage: ocf_stop_processes SIGNALS WAIT_TIME PIDS # # we send signals (use quotes for more than one!) in the order # given; if one or more processes are still running we try KILL; # the wait_time is the _total_ time we'll spend in this function # this time may be slightly exceeded if the processes won't leave # # returns: # 0: all processes left # 1: some processes still running # # example: # # ocf_stop_processes TERM 5 $pids # ocf_stop_processes() { local signals="$1" local wait_time="$(($2/`echo $signals|wc -w`))" shift 2 local pids="$*" local sig i test -z "$pids" && return 0 for sig in $signals KILL; do kill -s $sig $pids 2>/dev/null # try to leave early, and yet leave processes time to exit sleep 0.2 for i in `seq $wait_time`; do kill -s 0 $pids 2>/dev/null || return 0 sleep 1 done done return 1 } # # create a given status directory # if the directory path doesn't start with $HA_VARRUN, then # we return with error (most of the calls would be with the user # supplied configuration, hence we need to do necessary # protection) # used mostly for PID files # # usage: ocf_mkstatedir owner permissions path # # owner: user.group # permissions: permissions # path: directory path # # example: # ocf_mkstatedir named 755 `dirname $pidfile` # ocf_mkstatedir() { local owner local perms local path owner=$1 perms=$2 path=$3 test -d $path && return 0 [ $(id -u) = 0 ] || return 1 case $path in ${HA_VARRUN%/}/*) : this path is ok ;; *) ocf_log err "cannot create $path (does not start with $HA_VARRUN)" return 1 ;; esac mkdir -p $path && chown $owner $path && chmod $perms $path } # # create a unique status directory in $HA_VARRUN # used mostly for PID files # the directory is by default set to # $HA_VARRUN/$OCF_RESOURCE_INSTANCE # the directory name is printed to stdout # # usage: ocf_unique_rundir owner permissions name # # owner: user.group (default: "root") # permissions: permissions (default: "755") # name: some unique string (default: "$OCF_RESOURCE_INSTANCE") # # to use the default either don't set the parameter or set it to # empty string ("") # example: # # STATEDIR=`ocf_unique_rundir named "" myownstatedir` # ocf_unique_rundir() { local path local owner local perms local name owner=${1:-"root"} perms=${2:-"755"} name=${3:-"$OCF_RESOURCE_INSTANCE"} path=$HA_VARRUN/$name if [ ! -d $path ]; then [ $(id -u) = 0 ] || return 1 mkdir -p $path && chown $owner $path && chmod $perms $path || return 1 fi echo $path } # # RA tracing may be turned on by setting OCF_TRACE_RA # the trace output will be saved to OCF_TRACE_FILE, if set, or # by default to # $HA_VARLIB/trace_ra//.. # e.g. $HA_VARLIB/trace_ra/oracle/db.start.2012-11-27.08:37:08 # # OCF_TRACE_FILE: # - FD (small integer [3-9]) in that case it is up to the callers # to capture output; the FD _must_ be open for writing # - absolute path # # NB: FD 9 may be used for tracing with bash >= v4 in case # OCF_TRACE_FILE is set to a path. # ocf_bash_has_xtracefd() { [ -n "$BASH_VERSION" ] && [ ${BASH_VERSINFO[0]} -ge 4 ] } # for backwards compatibility ocf_is_bash4() { ocf_bash_has_xtracefd } ocf_trace_redirect_to_file() { local dest=$1 if ocf_bash_has_xtracefd; then exec 9>$dest BASH_XTRACEFD=9 else exec 2>$dest fi } ocf_trace_redirect_to_fd() { local fd=$1 if ocf_bash_has_xtracefd; then BASH_XTRACEFD=$fd else exec 2>&$fd fi } __ocf_test_trc_dest() { local dest=$1 if ! touch $dest; then ocf_log warn "$dest not writable, trace not going to happen" __OCF_TRC_DEST="" __OCF_TRC_MANAGE="" return 1 fi return 0 } ocf_default_trace_dest() { tty >/dev/null && return if [ -n "$OCF_RESOURCE_TYPE" -a \ -n "$OCF_RESOURCE_INSTANCE" -a -n "$__OCF_ACTION" ]; then local ts=`date +%F.%T` __OCF_TRC_DEST=${OCF_RESKEY_trace_dir}/${OCF_RESOURCE_TYPE}/${OCF_RESOURCE_INSTANCE}.${__OCF_ACTION}.$ts __OCF_TRC_MANAGE="1" fi } ocf_start_trace() { export __OCF_TRC_DEST="" __OCF_TRC_MANAGE="" case "$OCF_TRACE_FILE" in [3-9]) ocf_trace_redirect_to_fd "$OCF_TRACE_FILE" ;; /*/*) __OCF_TRC_DEST=$OCF_TRACE_FILE ;; "") ocf_default_trace_dest ;; *) ocf_log warn "OCF_TRACE_FILE must be set to either FD (open for writing) or absolute file path" ocf_default_trace_dest ;; esac if [ "$__OCF_TRC_DEST" ]; then mkdir -p `dirname $__OCF_TRC_DEST` __ocf_test_trc_dest $__OCF_TRC_DEST || return ocf_trace_redirect_to_file "$__OCF_TRC_DEST" fi if [ -n "$BASH_VERSION" ]; then PS4='+ `date +"%T"`: ${FUNCNAME[0]:+${FUNCNAME[0]}:}${LINENO}: ' fi set -x env=$( echo; printenv | sort ) } ocf_stop_trace() { set +x } # Helper functions to map from nodename/bundle-name and physical hostname # list_index_for_word "node0 node1 node2 node3 node4 node5" node4 --> 5 # list_word_at_index "NA host1 host2 host3 host4 host5" 3 --> host2 # list_index_for_word "node1 node2 node3 node4 node5" node7 --> "" # list_word_at_index "host1 host2 host3 host4 host5" 8 --> "" # attribute_target node1 --> host1 list_index_for_word() { echo $1 | tr ' ' '\n' | awk -v x="$2" '$0~x {print NR}' } list_word_at_index() { echo $1 | tr ' ' '\n' | awk -v n="$2" 'n == NR' } ocf_attribute_target() { if [ x$1 = x ]; then if [ x$OCF_RESKEY_CRM_meta_container_attribute_target = xhost -a x$OCF_RESKEY_CRM_meta_physical_host != x ]; then echo $OCF_RESKEY_CRM_meta_physical_host else if [ x$OCF_RESKEY_CRM_meta_on_node != x ]; then echo $OCF_RESKEY_CRM_meta_on_node else ocf_local_nodename fi fi return elif [ x"$OCF_RESKEY_CRM_meta_notify_all_uname" != x ]; then index=$(list_index_for_word "$OCF_RESKEY_CRM_meta_notify_all_uname" $1) mapping="" if [ x$index != x ]; then mapping=$(list_word_at_index "$OCF_RESKEY_CRM_meta_notify_all_hosts" $index) fi if [ x$mapping != x -a x$mapping != xNA ]; then echo $mapping return fi fi echo $1 } ocf_promotion_score() { ocf_version_cmp "$OCF_RESKEY_crm_feature_set" "3.10.0" res=$? if [ $res -eq 2 ] || [ $res -eq 1 ] || ! have_binary "crm_master"; then ${HA_SBIN_DIR}/crm_attribute -p ${OCF_RESOURCE_INSTANCE} $@ else ${HA_SBIN_DIR}/crm_master -l reboot $@ fi } __ocf_set_defaults "$@" : ${OCF_TRACE_RA:=$OCF_RESKEY_trace_ra} : ${OCF_RESKEY_trace_dir:="$HA_VARLIB/trace_ra"} ocf_is_true "$OCF_TRACE_RA" && ocf_start_trace # pacemaker sets HA_use_logd, some others use HA_LOGD :/ if ocf_is_true "$HA_use_logd"; then : ${HA_LOGD:=yes} fi diff --git a/heartbeat/rabbitmq-cluster.in b/heartbeat/rabbitmq-cluster.in index 0d36d95e9..169c7efaf 100755 --- a/heartbeat/rabbitmq-cluster.in +++ b/heartbeat/rabbitmq-cluster.in @@ -1,632 +1,647 @@ #!@BASH_SHELL@ # # Copyright (c) 2014 David Vossel # All Rights Reserved. # # This program is free software; you can redistribute it and/or modify # it under the terms of version 2 of the GNU General Public License as # published by the Free Software Foundation. # # This program is distributed in the hope that it would be useful, but # WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. # # Further, this software is distributed without any warranty that it is # free of the rightful claim of any third person regarding infringement # or the like. Any license provided herein, whether implied or # otherwise, applies only to this software file. Patent licenses, if # any, provided herein do not apply to combinations of this program with # other software, or any other product whatsoever. # # You should have received a copy of the GNU General Public License # along with this program; if not, write the Free Software Foundation, # Inc., 59 Temple Place - Suite 330, Boston MA 02111-1307, USA. # ####################################################################### # Initialization: : ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat} . ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs # Parameter defaults OCF_RESKEY_set_policy_default="" : ${OCF_RESKEY_set_policy=${OCF_RESKEY_set_policy_default}} ####################################################################### # This arbitrary value here is used by the rmq_start action to # signify that the resource agent must retry the start process # It might potentially conflict with OCF assigned error code # in the future. RMQ_TRY_RESTART_ERROR_CODE=126 RMQ_SERVER=/usr/sbin/rabbitmq-server RMQ_CTL=/usr/sbin/rabbitmqctl RMQ_EVAL="${RMQ_CTL} eval -q" RMQ_DATA_DIR="/var/lib/rabbitmq/mnesia" RMQ_PID_DIR="/var/run/rabbitmq" RMQ_PID_FILE="/var/run/rabbitmq/rmq.pid" RMQ_LOG_DIR="/var/log/rabbitmq" if [ "$__OCF_ACTION" != "meta-data" ]; then NODENAME=$(ocf_attribute_target) fi # this attr represents the current active local rmq node name. # when rmq stops or the node is fenced, this attr disappears RMQ_CRM_ATTR_COOKIE="rmq-node-attr-${OCF_RESOURCE_INSTANCE}" # this attr represents the last known active local rmq node name # when rmp stops or the node is fenced, the attr stays forever so # we can continue to map an offline pcmk node to it's rmq node name # equivalent. RMQ_CRM_ATTR_COOKIE_LAST_KNOWN="rmq-node-attr-last-known-${OCF_RESOURCE_INSTANCE}" meta_data() { cat < 1.0 Starts cloned rabbitmq cluster instance. NB: note that this RA cannot be spawned across a mix of pacemaker and pacemaker-remote nodes. Only on pacemaker *or* pacemaker-remote nodes exclusively. rabbitmq clustered Policy string to pass to 'rabbitmqctl set_policy' right after bootstrapping the first rabbitmq instance. rabbitmqctl set_policy args END } ####################################################################### rmq_usage() { cat < /dev/null 2>&1 } rmq_restore_users_perms_policies() { # Restore users, user permissions, and policies (if any) BaseDataDir=`dirname $RMQ_DATA_DIR` $RMQ_EVAL " %% Run only if Mnesia is ready. lists:any(fun({mnesia,_,_}) -> true; ({_,_,_}) -> false end, application:which_applications()) andalso begin Restore = fun(Table, PostprocessFun, Filename) -> case file:consult(Filename) of {error, _} -> ok; {ok, [Result]} -> lists:foreach(fun(X) -> mnesia:dirty_write(Table, PostprocessFun(X)) end, Result), file:delete(Filename) end end, %% Restore users Upgrade = fun ({internal_user, A, B, C}) -> {internal_user, A, B, C, rabbit_password_hashing_md5}; ({internal_user, A, B, C, D}) -> {internal_user, A, B, C, D} end, Downgrade = fun ({internal_user, A, B, C}) -> {internal_user, A, B, C}; ({internal_user, A, B, C, rabbit_password_hashing_md5}) -> {internal_user, A, B, C}; %% Incompatible scheme, so we will loose user's password ('B' value) during conversion. %% Unfortunately, this case will require manual intervention - user have to run: %% rabbitmqctl change_password ({internal_user, A, B, C, _}) -> {internal_user, A, B, C} end, %% Check db scheme first [WildPattern] = ets:select(mnesia_gvar, [ { {{rabbit_user, wild_pattern}, '\\\$1'}, [], ['\\\$1'] } ]), case WildPattern of %% Version < 3.6.0 {internal_user,'_','_','_'} -> Restore(rabbit_user, Downgrade, \"$BaseDataDir/users.erl\"); %% Version >= 3.6.0 {internal_user,'_','_','_','_'} -> Restore(rabbit_user, Upgrade, \"$BaseDataDir/users.erl\") end, NoOp = fun(X) -> X end, %% Restore user permissions Restore(rabbit_user_permission, NoOp, \"$BaseDataDir/users_perms.erl\"), %% Restore policies Restore(rabbit_runtime_parameters, NoOp, \"$BaseDataDir/policies.erl\") end. " } rmq_local_node() { local node_name=$($RMQ_CTL status 2>&1 | sed -n -e "s/^.*[S|s]tatus of node \(.*\)\s.*$/\1/p" | tr -d "'") if [ -z "$node_name" ]; then node_name=$(cat /etc/rabbitmq/rabbitmq-env.conf 2>/dev/null | grep "\s*RABBITMQ_NODENAME=" | awk -F= '{print $2}') fi echo "$node_name" } rmq_join_list() { local join_list=$(cibadmin -Q --xpath "//node_state[@crmd='online']//nvpair[@name='$RMQ_CRM_ATTR_COOKIE']" | grep "$RMQ_CRM_ATTR_COOKIE" | sed -n -e "s/^.*value=.\(.*\)\".*$/\1/p") # If join_list is empty we want to check if there are any remote nodes # where rabbitmq is allowed to run (i.e. nodes without the crmd=online selector) if [ -z "$join_list" ]; then # Get all the nodes written in the ATTR_COOKIE no matter if # they are online or not. This will be one line per node like # rabbit@overcloud-rabbit-0 # rabbit@overcloud-rabbit-1 # ... local remote_join_list=$(cibadmin -Q --xpath "//node_state//nvpair[@name='$RMQ_CRM_ATTR_COOKIE']" | grep "$RMQ_CRM_ATTR_COOKIE" | sed -n -e "s/^.*value=.\(.*\)\".*$/\1/p") # The following expression prepares a filter like '-e overcloud-rabbit-0 -e overcloud-rabbit-1 -e ...' - local filter=$(crm_mon_no_validation -r --as-xml | xmllint --format --xpath "//nodes//node[@online='true' and @standby='false']/@name" - | xargs -n1 echo | awk -F= '{print "-e "$2}') + local filter=$(crm_mon_no_validation -r $XMLOPT | xmllint --format --xpath "//nodes//node[@online='true' and @standby='false']/@name" - | xargs -n1 echo | awk -F= '{print "-e "$2}') # export the intersection which gives us only the nodes that # a) wrote their namein the cib attrd # b) run on nodes where pacemaker_remote is enabled join_list="$(echo $remote_join_list | grep $filter)" fi echo $join_list } rmq_write_nodename() { local node_name=$(rmq_local_node) if [ -z "$node_name" ]; then ocf_log err "Failed to determine rabbitmq node name, exiting" exit $OCF_ERR_GENERIC fi # store the pcmknode to rmq node mapping as a transient attribute. This allows # us to retrieve the join list with a simple xpath. ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "$RMQ_CRM_ATTR_COOKIE" -v "$node_name" # the pcmknode to rmq node mapping as a permanent attribute as well. this lets # us continue to map offline nodes to their equivalent rmq node name ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l forever --name "$RMQ_CRM_ATTR_COOKIE_LAST_KNOWN" -v "$node_name" } rmq_delete_nodename() { # remove node-name ${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l reboot --name "$RMQ_CRM_ATTR_COOKIE" -D } prepare_dir () { if [ ! -d ${1} ] ; then mkdir -p ${1} chown -R rabbitmq:rabbitmq ${1} chmod 755 ${1} fi } remove_pid () { rm -f ${RMQ_PID_FILE} > /dev/null 2>&1 } rmq_app_running() { if $RMQ_EVAL 'application:which_applications().' | grep -q '{rabbit,'; then ocf_log debug "RabbitMQ application is running" return $OCF_SUCCESS else ocf_log debug "RabbitMQ application is stopped" rmq_delete_nodename return $OCF_NOT_RUNNING fi } rmq_node_alive() { if $RMQ_EVAL 'ok.'; then ocf_log debug "RabbitMQ node is alive" return $OCF_SUCCESS else ocf_log debug "RabbitMQ node is down" rmq_delete_nodename return $OCF_NOT_RUNNING fi } rmq_monitor() { local rc status=$($RMQ_EVAL 'rabbit_mnesia:cluster_status_from_mnesia().' 2>&1) if echo "${status}" | grep -q '^{ok'; then pcs_running=$(rmq_join_list | wc -w) ocf_log debug "Pacemaker thinks ${pcs_running} RabbitMQ nodes are running" rmq_running=$($RMQ_EVAL 'length(mnesia:system_info(running_db_nodes)).') ocf_log debug "RabbitMQ thinks ${rmq_running} RabbitMQ nodes are running" if [ $(( $rmq_running * 2 )) -lt $pcs_running ]; then ocf_log info "RabbitMQ is a minority partition, failing monitor" rmq_delete_nodename return $OCF_ERR_GENERIC fi ocf_log debug "RabbitMQ server is running normally" rmq_write_nodename return $OCF_SUCCESS else ocf_log info "RabbitMQ server could not get cluster status from mnesia" ocf_log debug "${status}" rmq_delete_nodename return $OCF_NOT_RUNNING fi } rmq_init_and_wait() { local rc local wait_timeout local timeout_string prepare_dir $RMQ_PID_DIR prepare_dir $RMQ_LOG_DIR remove_pid # the server startup script uses this environment variable export RABBITMQ_PID_FILE="$RMQ_PID_FILE" setsid sh -c "$RMQ_SERVER > ${RMQ_LOG_DIR}/startup_log 2> ${RMQ_LOG_DIR}/startup_err" & ocf_log info "Waiting for server to start" # We want to give the wait command almost the full startup timeout we are given # So we use the start operation timeout (in ms), convert it and subtract 5 seconds # In the silly case that it is less than 10 seconds we just skip setting the timeout wait_timeout=`expr $OCF_RESKEY_CRM_meta_timeout / 1000 - 5` if [ $wait_timeout -gt 10 ]; then timeout_string="--timeout ${wait_timeout}" else timeout_string="" fi $RMQ_CTL $timeout_string wait $RMQ_PID_FILE rc=$? if [ $rc -ne $OCF_SUCCESS ]; then remove_pid ocf_log info "rabbitmq-server start failed with a timeout of ($timeout_string): $rc" return $OCF_ERR_GENERIC fi rmq_app_running return $? } rmq_set_policy() { $RMQ_CTL set_policy "$@" > /dev/null 2>&1 } rmq_start_first() { local rc ocf_log info "Bootstrapping rabbitmq cluster" rmq_wipe_data rmq_init_and_wait rc=$? if [ $rc -eq 0 ]; then rc=$OCF_SUCCESS ocf_log info "cluster bootstrapped" rmq_write_nodename if [ -n "$OCF_RESKEY_set_policy" ]; then # do not quote set_policy, we are passing in arguments rmq_set_policy $OCF_RESKEY_set_policy if [ $? -ne 0 ]; then ocf_log err "Failed to set policy: $OCF_RESKEY_set_policy" rc=$OCF_ERR_GENERIC else ocf_log info "Policy set: $OCF_RESKEY_set_policy" fi fi else ocf_log info "failed to bootstrap cluster. Check SELINUX policy" rc=$OCF_ERR_GENERIC fi return $rc } rmq_is_clustered() { $RMQ_EVAL 'rabbit_mnesia:is_clustered().' | grep -q true } rmq_join_existing() { local join_list="$1" local rc=$OCF_ERR_GENERIC ocf_log info "Joining existing cluster with [ $(echo $join_list | tr '\n' ' ') ] nodes." rmq_init_and_wait if [ $? -ne 0 ]; then return $OCF_ERR_GENERIC fi if rmq_is_clustered; then ocf_log info "Successfully re-joined existing rabbitmq cluster automatically" return $OCF_SUCCESS fi # unconditionally join the cluster $RMQ_CTL stop_app > /dev/null 2>&1 for node in $(echo "$join_list"); do ocf_log info "Attempting to join cluster with target node $node" $RMQ_CTL join_cluster $node if [ $? -eq 0 ]; then ocf_log info "Joined cluster by connecting to node $node, starting app" $RMQ_CTL start_app rc=$? if [ $rc -ne 0 ]; then ocf_log err "'$RMQ_CTL start_app' failed" fi break; fi done if [ "$rc" -ne 0 ]; then ocf_log info "Join process incomplete, shutting down." return $OCF_ERR_GENERIC fi ocf_log info "Successfully joined existing rabbitmq cluster" return $OCF_SUCCESS } rmq_forget_cluster_node_remotely() { local running_cluster_nodes="$1" local node_to_forget="$2" ocf_log info "Forgetting $node_to_forget via nodes [ $(echo $running_cluster_nodes | tr '\n' ' ') ]." for running_cluster_node in $running_cluster_nodes; do $RMQ_CTL -n $running_cluster_node forget_cluster_node $node_to_forget if [ $? = 0 ]; then ocf_log info "Succeeded forgetting $node_to_forget via $running_cluster_node." return else ocf_log err "Failed to forget node $node_to_forget via $running_cluster_node." fi done } rmq_notify() { node_list="${OCF_RESKEY_CRM_meta_notify_stop_uname}" mode="${OCF_RESKEY_CRM_meta_notify_type}-${OCF_RESKEY_CRM_meta_notify_operation}" # When notifications are on, this agent is going to "forget" nodes once they # leave the cluster. This is thought to resolve some issues where rabbitmq # blocks trying to sync with an offline node after a fencing action occurs. if ! [ "${mode}" = "post-stop" ]; then return $OCF_SUCCESS fi rmq_monitor if [ $? -ne $OCF_SUCCESS ]; then # only run forget when we are for sure active return $OCF_SUCCESS fi # forget each stopped rmq instance in the provided pcmk node in the list. for node in $(echo "$node_list"); do local rmq_node="$(${HA_SBIN_DIR}/crm_attribute -N $(ocf_attribute_target $node) -l forever --query --name $RMQ_CRM_ATTR_COOKIE_LAST_KNOWN -q)" if [ -z "$rmq_node" ]; then ocf_log warn "Unable to map pcmk node $node to a known rmq node." continue fi ocf_log notice "Forgetting stopped node $rmq_node" $RMQ_CTL forget_cluster_node $rmq_node if [ $? -ne 0 ]; then ocf_log warn "Unable to forget offline node $rmq_node." fi done return $OCF_SUCCESS } rmq_try_start() { local join_list="" local rc rmq_monitor if [ $? -eq $OCF_SUCCESS ]; then return $OCF_SUCCESS fi join_list=$(rmq_join_list) # No join list means no active instances are up. This instance # is the first, so it needs to bootstrap the rest if [ -z "$join_list" ]; then rmq_start_first rc=$? rmq_restore_users_perms_policies return $rc fi # Try to join existing cluster ocf_log info "wiping data directory before joining" local local_rmq_node="$(${HA_SBIN_DIR}/crm_attribute -N $NODENAME -l forever --query --name $RMQ_CRM_ATTR_COOKIE_LAST_KNOWN -q)" rmq_stop rmq_wipe_data if [ -z "$local_rmq_node" ]; then ocf_log warn "Unable to forget the cluster node because local node name cannot be detected" else rmq_forget_cluster_node_remotely "$join_list" "$local_rmq_node" fi rmq_join_existing "$join_list" rc=$? if [ $rc -ne 0 ]; then # we could not join the rabbitmq cluster from any of the running nodes # this might be due to a unexpected reset of those nodes. Give ourself # a chance to start by retrying the entire start sequence. ocf_log warn "Failed to join the RabbitMQ cluster from nodes ${join_list}. Stopping local unclustered rabbitmq" rmq_stop ocf_log warn "Re-detect available rabbitmq nodes and try to start again" # return an unused OCF value to signify a "retry" condition return $RMQ_TRY_RESTART_ERROR_CODE fi rmq_restore_users_perms_policies return $OCF_SUCCESS } rmq_start() { local rc=$RMQ_TRY_RESTART_ERROR_CODE while [ $rc -eq $RMQ_TRY_RESTART_ERROR_CODE ]; do rmq_try_start rc=$? done return $rc } rmq_stop() { # Backup users, user permissions, and policies BaseDataDir=`dirname $RMQ_DATA_DIR` $RMQ_EVAL " %% Run only if Mnesia is still available. lists:any(fun({mnesia,_,_}) -> true; ({_,_,_}) -> false end, application:which_applications()) andalso begin Backup = fun(Table, SelectPattern, Filter, Filename) -> Result = case catch mnesia:dirty_select(Table, [{SelectPattern, [Filter], ['\\\$_']}]) of {'EXIT', _} -> []; Any -> Any end, Result /= [] andalso file:write_file(Filename, io_lib:fwrite(\"~p.~n\", [Result])) end, %% Backup users %% Check db scheme first [WildPattern] = ets:select(mnesia_gvar, [ { {{rabbit_user, wild_pattern}, '\\\$1'}, [], ['\\\$1'] } ]), UsersSelectPattern = case WildPattern of %% Version < 3.6.0 {internal_user,'_','_','_'} -> {internal_user, '\\\$1', '_', '_'}; %% Version >= 3.6.0 {internal_user,'_','_','_','_'} -> {internal_user, '\\\$1', '_', '_', '_'} end, Backup(rabbit_user, UsersSelectPattern, {'/=', '\\\$1', <<\"guest\">>}, \"$BaseDataDir/users.erl\"), %% Backup user permissions Backup(rabbit_user_permission, {'\\\$1', {'\\\$2', '\\\$3','\\\$4'}, '\\\$5'}, {'/=', '\\\$3', <<\"guest\">>}, \"$BaseDataDir/users_perms.erl\"), %% Backup policies Backup(rabbit_runtime_parameters, {runtime_parameters, {'_', '\\\$1', '_'}, '_'}, {'==', '\\\$1', <<\"policy\">>}, \"$BaseDataDir/policies.erl\") end. " rmq_node_alive if [ $? -eq $OCF_NOT_RUNNING ]; then return $OCF_SUCCESS fi $RMQ_CTL stop rc=$? if [ $rc -ne 0 ]; then ocf_log err "rabbitmq-server stop command failed: $RMQ_CTL stop, $rc" rmq_delete_nodename return $rc fi #TODO add kill logic stop_wait=1 while [ $stop_wait = 1 ]; do rmq_app_running rc=$? if [ "$rc" -eq $OCF_NOT_RUNNING ]; then stop_wait=0 break elif [ "$rc" -ne $OCF_SUCCESS ]; then ocf_log info "rabbitmq-server stop failed: $rc" rmq_delete_nodename exit $OCF_ERR_GENERIC fi sleep 1 done rmq_delete_nodename remove_pid return $OCF_SUCCESS } rmq_validate() { check_binary $RMQ_SERVER check_binary $RMQ_CTL # This resource only makes sense as a clone right now. at some point # we may want to verify the following. #TODO verify cloned #TODO verify ordered=true # Given that this resource does the cluster join explicitly, # having a cluster_nodes list in the static config file will # likely conflict with this agent. #TODO verify no cluster list in rabbitmq conf #cat /etc/rabbitmq/rabbitmq.config | grep "cluster_nodes" return $OCF_SUCCESS } +ocf_version_cmp "$OCF_RESKEY_crm_feature_set" "3.1.0" +res=$? +if [ -z "$OCF_RESKEY_crm_feature_set" ] || [ $res -eq 2 ]; then + XMLOPT="--output-as=xml" + ocf_version_cmp "$OCF_RESKEY_crm_feature_set" "3.2.0" + if [ $? -eq 1 ]; then + crm_mon_no_validation -1 $XMLOPT >/dev/null 2>&1 + if [ $? -ne 0 ]; then + XMLOPT="--as-xml" + fi + fi +else + XMLOPT="--as-xml" +fi + case $__OCF_ACTION in meta-data) meta_data exit $OCF_SUCCESS ;; start) rmq_start;; stop) rmq_stop;; monitor) rmq_monitor;; validate-all) rmq_validate;; notify) rmq_notify;; usage|help) rmq_usage exit $OCF_SUCCESS ;; *) rmq_usage exit $OCF_ERR_UNIMPLEMENTED ;; esac rc=$? ocf_log debug "${OCF_RESOURCE_INSTANCE} $__OCF_ACTION : $rc" exit $rc