diff options
Diffstat (limited to 'extras/ganesha/ocf/ganesha_grace')
| -rw-r--r-- | extras/ganesha/ocf/ganesha_grace | 168 |
1 files changed, 110 insertions, 58 deletions
diff --git a/extras/ganesha/ocf/ganesha_grace b/extras/ganesha/ocf/ganesha_grace index 397b0abdeab..825f7164597 100644 --- a/extras/ganesha/ocf/ganesha_grace +++ b/extras/ganesha/ocf/ganesha_grace @@ -30,14 +30,17 @@ . ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs if [ -n "$OCF_DEBUG_LIBRARY" ]; then - . $OCF_DEBUG_LIBRARY + . $OCF_DEBUG_LIBRARY else - : ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat} -. ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs + : ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat} + . ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs fi +OCF_RESKEY_grace_active_default="grace-active" +: ${OCF_RESKEY_grace_active=${OCF_RESKEY_grace_active_default}} + ganesha_meta_data() { - cat <<END + cat <<END <?xml version="1.0"?> <!DOCTYPE resource-agent SYSTEM "ra-api-1.dtd"> <resource-agent name="ganesha_grace"> @@ -51,19 +54,25 @@ resource agent for nfs-ganesha. <shortdesc lang="en">Manages the user-space nfs-ganesha NFS server</shortdesc> <parameters> +<parameter name="grace_active"> +<longdesc lang="en">NFS-Ganesha grace active attribute</longdesc> +<shortdesc lang="en">NFS-Ganesha grace active attribute</shortdesc> +<content type="string" default="grace-active" /> +</parameter> </parameters> <actions> <action name="start" timeout="40s" /> <action name="stop" timeout="40s" /> -<action name="status" depth="0" timeout="20s" interval="5s" /> -<action name="monitor" depth="0" timeout="20s" interval="5s" /> +<action name="status" timeout="20s" interval="60s" /> +<action name="monitor" depth="0" timeout="10s" interval="5s" /> +<action name="notify" timeout="10s" /> <action name="meta-data" timeout="20s" /> </actions> </resource-agent> END -return $OCF_SUCCESS +return ${OCF_SUCCESS} } ganesha_grace_usage() { @@ -73,10 +82,10 @@ ganesha_grace_usage() { # Make sure meta-data and usage always succeed case $__OCF_ACTION in meta-data) ganesha_meta_data - exit $OCF_SUCCESS + exit ${OCF_SUCCESS} ;; usage|help) ganesha_usage - exit $OCF_SUCCESS + exit ${OCF_SUCCESS} ;; *) ;; @@ -84,70 +93,112 @@ esac ganesha_grace_start() { - local result="" - local resourcename="" - local deadserver="" - local tmpIFS=${IFS} - - # logger "ganesha_grace_start()" - # we're here because somewhere in the cluster one or more - # of the ganesha.nfsds have died, triggering a floating IP - # address to move. Resource constraint location rules ensure - # that this is invoked before the floating IP is moved. - if [ -e /var/run/ganesha.nfsd.pid -a \ - -d /proc/$(cat /var/run/ganesha.nfsd.pid) ]; then - # my ganesha.nfsd is still running - # find out which one died? - - pcs status | grep dead_ip-1 | sort > /tmp/.pcs_status - - result=$(diff /var/run/ganesha/pcs_status /tmp/.pcs_status | grep '^>') - if [[ ${result} ]]; then - # logger "ganesha_grace_start(), ${result}" - IFS=$'\n' - for line in ${result}; do - resourcename=$(echo ${line} | cut -f 1 | cut -d ' ' -f 3) - deadserver=${resourcename%"-dead_ip-1"} - - if [[ ${deadserver} ]]; then - # logger "ganesha_grace_start(), ${line}" - # logger "ganesha_grace_start(), dbus-send --print-reply --system --dest=org.ganesha.nfsd /org/ganesha/nfsd/admin org.ganesha.nfsd.admin.grace string:${deadserver}" - dbus-send --print-reply --system --dest=org.ganesha.nfsd /org/ganesha/nfsd/admin org.ganesha.nfsd.admin.grace string:${deadserver} - if [ $? -ne 0 ]; then - logger "warning: dbus-send --print-reply --system --dest=org.ganesha.nfsd /org/ganesha/nfsd/admin org.ganesha.nfsd.admin.grace string:${deadserver} failed" - fi - fi - done - IFS=${tmpIFS} - fi + local rc=${OCF_ERR_GENERIC} + local host=$(hostname -s) + + ocf_log debug "ganesha_grace_start()" + # give ganesha_mon RA a chance to set the crm_attr first + # I mislike the sleep, but it's not clear that looping + # with a small sleep is necessarily better + # start has a 40sec timeout, so a 5sec sleep here is okay + sleep 5 + attr=$(crm_attribute --query --node=${host} --name=${OCF_RESKEY_grace_active} 2> /dev/null) + if [ $? -ne 0 ]; then + host=$(hostname) + attr=$(crm_attribute --query --node=${host} --name=${OCF_RESKEY_grace_active} 2> /dev/null ) + if [ $? -ne 0 ]; then + ocf_log info "grace start: crm_attribute --query --node=${host} --name=${OCF_RESKEY_grace_active} failed" + fi + fi + + # Three possibilities: + # 1. There is no attribute at all and attr_updater returns + # a zero length string. This happens when + # ganesha_mon::monitor hasn't run at least once to set + # the attribute. The assumption here is that the system + # is coming up. We pretend, for now, that the node is + # healthy, to allow the system to continue coming up. + # It will cure itself in a few seconds + # 2. There is an attribute, and it has the value "1"; this + # node is healthy. + # 3. There is an attribute, but it has no value or the value + # "0"; this node is not healthy. + + # case 1 + if [[ -z "${attr}" ]]; then + return ${OCF_SUCCESS} + fi + # case 2 + if [[ "${attr}" = *"value=1" ]]; then + return ${OCF_SUCCESS} fi - return $OCF_SUCCESS + + # case 3 + return ${OCF_NOT_RUNNING} } ganesha_grace_stop() { - # logger "ganesha_grace_stop()" - return $OCF_SUCCESS + ocf_log debug "ganesha_grace_stop()" + return ${OCF_SUCCESS} +} + +ganesha_grace_notify() +{ + # since this is a clone RA we should only ever see pre-start + # or post-stop + mode="${OCF_RESKEY_CRM_meta_notify_type}-${OCF_RESKEY_CRM_meta_notify_operation}" + case "${mode}" in + pre-start | post-stop) + dbus-send --print-reply --system --dest=org.ganesha.nfsd /org/ganesha/nfsd/admin org.ganesha.nfsd.admin.grace string:${OCF_RESKEY_CRM_meta_notify_stop_uname} + if [ $? -ne 0 ]; then + ocf_log info "dbus-send --print-reply --system --dest=org.ganesha.nfsd /org/ganesha/nfsd/admin org.ganesha.nfsd.admin.grace string:${OCF_RESKEY_CRM_meta_notify_stop_uname} failed" + fi + ;; + esac + + return ${OCF_SUCCESS} } ganesha_grace_monitor() { - # logger "ganesha_grace_monitor()" - pcs status | grep dead_ip-1 | sort > /var/run/ganesha/pcs_status - return $OCF_SUCCESS + local host=$(hostname -s) + + ocf_log debug "monitor" + + attr=$(crm_attribute --query --node=${host} --name=${OCF_RESKEY_grace_active} 2> /dev/null) + if [ $? -ne 0 ]; then + host=$(hostname) + attr=$(crm_attribute --query --node=${host} --name=${OCF_RESKEY_grace_active} 2> /dev/null) + if [ $? -ne 0 ]; then + ocf_log info "crm_attribute --query --node=${host} --name=${OCF_RESKEY_grace_active} failed" + fi + fi + + # if there is no attribute (yet), maybe it's because + # this RA started before ganesha_mon (nfs-mon) has had + # chance to create it. In which case we'll pretend + # everything is okay this time around + if [[ -z "${attr}" ]]; then + return ${OCF_SUCCESS} + fi + + if [[ "${attr}" = *"value=1" ]]; then + return ${OCF_SUCCESS} + fi + + return ${OCF_NOT_RUNNING} } ganesha_grace_validate() { - return $OCF_SUCCESS + return ${OCF_SUCCESS} } ganesha_grace_validate -# logger "ganesha_grace ${OCF_RESOURCE_INSTANCE} $__OCF_ACTION" - # Translate each action into the appropriate function call case $__OCF_ACTION in start) ganesha_grace_start @@ -156,14 +207,15 @@ stop) ganesha_grace_stop ;; status|monitor) ganesha_grace_monitor ;; +notify) ganesha_grace_notify + ;; *) ganesha_grace_usage - exit $OCF_ERR_UNIMPLEMENTED - ;; + exit ${OCF_ERR_UNIMPLEMENTED} + ;; esac rc=$? # The resource agent may optionally log a debug message -ocf_log debug "${OCF_RESOURCE_INSTANCE} $__OCF_ACTION returned $rc" +ocf_log debug "${OCF_RESOURCE_INSTANCE} ${__OCF_ACTION} returned $rc" exit $rc - |