diff --git a/heartbeat/Filesystem b/heartbeat/Filesystem index b48bee142..68b061233 100755 --- a/heartbeat/Filesystem +++ b/heartbeat/Filesystem @@ -1,1216 +1,1225 @@ #!/bin/sh # # Support: users@clusterlabs.org # License: GNU General Public License (GPL) # # Filesystem # Description: Manages a Filesystem on a shared storage medium. # Original Author: Eric Z. Ayers (eric.ayers@compgen.com) # Original Release: 25 Oct 2000 # # usage: ./Filesystem {start|stop|status|monitor|validate-all|meta-data} # # OCF parameters are as below: # OCF_RESKEY_device # OCF_RESKEY_directory # OCF_RESKEY_fstype # OCF_RESKEY_options # OCF_RESKEY_statusfile_prefix # OCF_RESKEY_run_fsck # OCF_RESKEY_fast_stop # OCF_RESKEY_force_clones # #OCF_RESKEY_device : name of block device for the filesystem. e.g. /dev/sda1, /dev/md0 # Or a -U or -L option for mount, or an NFS mount specification #OCF_RESKEY_directory : the mount point for the filesystem #OCF_RESKEY_fstype : optional name of the filesystem type. e.g. ext2 #OCF_RESKEY_options : options to be given to the mount command via -o #OCF_RESKEY_statusfile_prefix : the prefix used for a status file for monitoring #OCF_RESKEY_run_fsck : fsck execution mode: auto(default)/force/no #OCF_RESKEY_fast_stop : fast stop: yes(default)/no #OCF_RESKEY_force_clones : allow running the resource as clone. e.g. local xfs mounts # for each brick in a glusterfs setup # # # This assumes you want to manage a filesystem on a shared (SCSI) bus, # on a replicated device (such as DRBD), or a network filesystem (such # as NFS or Samba). # # Do not put this filesystem in /etc/fstab. This script manages all of # that for you. # # NOTE: If 2 or more nodes mount the same file system read-write, and # that file system is not designed for that specific purpose # (such as GFS or OCFS2), and is not a network file system like # NFS or Samba, then the filesystem is going to become # corrupted. # # As a result, you should use this together with the stonith # option and redundant, independent communications paths. # # If you don't do this, don't blame us when you scramble your # disk. ####################################################################### # Initialization: : ${OCF_FUNCTIONS_DIR=${OCF_ROOT}/lib/heartbeat} . ${OCF_FUNCTIONS_DIR}/ocf-shellfuncs # Defaults DFLT_STATUSDIR=".Filesystem_status/" # Parameter defaults OCF_RESKEY_device_default="" OCF_RESKEY_directory_default="" OCF_RESKEY_fstype_default="" OCF_RESKEY_options_default="" OCF_RESKEY_statusfile_prefix_default="${DFLT_STATUSDIR}" OCF_RESKEY_run_fsck_default="auto" OCF_RESKEY_fast_stop_default="no" OCF_RESKEY_force_clones_default="false" OCF_RESKEY_force_unmount_default="true" OCF_RESKEY_term_signals_default="TERM" OCF_RESKEY_kill_signals_default="KILL" OCF_RESKEY_signal_delay_default="1" # RHEL specific defaults if is_redhat_based; then get_os_ver ocf_version_cmp "$VER" "9.0" 2>/dev/null case "$?" in # RHEL >= 9 1|2) OCF_RESKEY_force_unmount_default="safe";; # RHEL < 9 and fallback if ocf_version_cmp() fails *) OCF_RESKEY_fast_stop_default="yes";; esac fi : ${OCF_RESKEY_device=${OCF_RESKEY_device_default}} : ${OCF_RESKEY_directory=${OCF_RESKEY_directory_default}} : ${OCF_RESKEY_fstype=${OCF_RESKEY_fstype_default}} : ${OCF_RESKEY_options=${OCF_RESKEY_options_default}} : ${OCF_RESKEY_statusfile_prefix=${OCF_RESKEY_statusfile_prefix_default}} : ${OCF_RESKEY_run_fsck=${OCF_RESKEY_run_fsck_default}} if [ -z "${OCF_RESKEY_fast_stop}" ]; then case "$OCF_RESKEY_fstype" in gfs2) OCF_RESKEY_fast_stop="no";; *) OCF_RESKEY_fast_stop=${OCF_RESKEY_fast_stop_default};; esac fi : ${OCF_RESKEY_force_clones=${OCF_RESKEY_force_clones_default}} : ${OCF_RESKEY_force_unmount=${OCF_RESKEY_force_unmount_default}} : ${OCF_RESKEY_term_signals=${OCF_RESKEY_term_signals_default}} : ${OCF_RESKEY_kill_signals=${OCF_RESKEY_kill_signals_default}} : ${OCF_RESKEY_signal_delay=${OCF_RESKEY_signal_delay_default}} # Variables used by multiple methods HOSTOS=$(uname) TAB=' ' # The status file is going to an extra directory, by default # prefix=${OCF_RESKEY_statusfile_prefix} : ${prefix:=$DFLT_STATUSDIR} suffix="${OCF_RESOURCE_INSTANCE}" [ "$OCF_RESKEY_CRM_meta_clone" ] && suffix="${suffix}_$OCF_RESKEY_CRM_meta_clone" suffix="${suffix}_$(uname -n)" STATUSFILE="${OCF_RESKEY_directory}/$prefix$suffix" ####################################################################### usage() { cat <<-EOT usage: $0 {start|stop|status|monitor|validate-all|meta-data} EOT } meta_data() { cat < 1.0 Resource script for Filesystem. It manages a Filesystem on a shared storage medium. The standard monitor operation of depth 0 (also known as probe) checks if the filesystem is mounted. If you want deeper tests, set OCF_CHECK_LEVEL to one of the following values: 10: read first 16 blocks of the device (raw read) This doesn't exercise the filesystem at all, but the device on which the filesystem lives. This is noop for non-block devices such as NFS, SMBFS, or bind mounts. 20: test if a status file can be written and read The status file must be writable by root. This is not always the case with an NFS mount, as NFS exports usually have the "root_squash" option set. In such a setup, you must either use read-only monitoring (depth=10), export with "no_root_squash" on your NFS server, or grant world write permissions on the directory where the status file is to be placed. Manages filesystem mounts The name of block device for the filesystem, or -U, -L options for mount, or NFS mount specification. NOTE: On Linux /dev/disk/by-{uuid,label}/ are preferred to -U/-L. block device The mount point for the filesystem. mount point The type of filesystem to be mounted. filesystem type Any extra options to be given as -o options to mount. For bind mounts, add "bind" here and set fstype to "none". We will do the right thing for options such as "bind,ro". options The prefix to be used for a status file for resource monitoring with depth 20. If you don't specify this parameter, all status files will be created in a separate directory. status file prefix Specify how to decide whether to run fsck or not. "auto" : decide to run fsck depending on the fstype(default) "force" : always run fsck regardless of the fstype "no" : do not run fsck ever. run_fsck Normally, we expect no users of the filesystem and the stop operation to finish quickly. If you cannot control the filesystem users easily and want to prevent the stop action from failing, then set this parameter to "no" and add an appropriate timeout for the stop operation. This defaults to "no" for GFS2 filesystems. fast stop The use of a clone setup for local filesystems is forbidden by default. For special setups like glusterfs, cloning a mount of a local device with a filesystem like ext4 or xfs independently on several nodes is a valid use case. Only set this to "true" if you know what you are doing! allow running as a clone, regardless of filesystem type This option allows specifying how to handle processes that are currently accessing the mount directory. "true" : Kill processes accessing mount point "safe" : Kill processes accessing mount point using methods that avoid functions that could potentially block during process detection "false" : Do not kill any processes. The 'safe' option uses shell logic to walk the /procs/ directory for pids using the mount point while the default option uses the fuser cli tool. fuser is known to perform operations that can potentially block if unresponsive nfs mounts are in use on the system. Kill processes before unmount Signals (names or numbers, whitespace separated) to send processes during graceful termination phase in stop-action. Signals (names or numbers, whitespace separated) to send processes during graceful termination phase in stop-action Signals (names or numbers, whitespace separated) to send processes during forceful killing phase in stop-action. Signals (names or numbers, whitespace separated) to send processes during forceful killing phase in stop-action How many seconds to wait after sending term/kill signals to processes in stop-action. How many seconds to wait after sending term/kill signals to processes in stop-action END } # # Make sure the kernel does the right thing with the FS buffers # This function should be called after unmounting and before mounting # It may not be necessary in 2.4 and later kernels, but it shouldn't hurt # anything either... # # It's really a bug that you have to do this at all... # flushbufs() { if have_binary $BLOCKDEV ; then if [ "$blockdevice" = "yes" ] ; then $BLOCKDEV --flushbufs $1 return $? fi fi return 0 } # Take advantage of /etc/mtab if present, use portable mount command # otherwise. Normalize format to "dev mountpoint fstype". is_bind_mount() { echo "$options" | grep -w bind >/dev/null 2>&1 } list_mounts() { local inpf="" local mount_list="" local check_list="x" if [ -e "/proc/mounts" ] && ! is_bind_mount; then inpf=/proc/mounts elif [ -f "/etc/mtab" -a -r "/etc/mtab" ]; then inpf=/etc/mtab fi # Make sure that the mount list has not been changed while reading. while [ "$mount_list" != "$check_list" ]; do check_list="$mount_list" if [ "$inpf" ]; then # ... # Spaces in device or mountpoint are octal \040 in $inpf # Convert literal spaces (field separators) to tabs mount_list=$(cut -d' ' -f1,2,3 < $inpf | tr ' ' "$TAB") else # on type ... # Use tabs as field separators match_string='\(.*\) on \(.*\) type \([^[:space:]]\+\) .*' replace_string="\\1${TAB}\\2${TAB}\\3" mount_list=$($MOUNT | sed "s/$match_string/$replace_string/g") fi done # Convert octal \040 to space characters printf "$mount_list" } determine_blockdevice() { if [ $blockdevice = "yes" ]; then return fi # Get the current real device name, if possible. # (specified devname could be -L or -U...) case "$FSTYPE" in nfs4|nfs|aznfs|efs|smbfs|cifs|glusterfs|ceph|tmpfs|overlay|overlayfs|rozofs|zfs|cvfs|none|lustre) : ;; *) match_string="${TAB}${CANONICALIZED_MOUNTPOINT}${TAB}" DEVICE=$(list_mounts | grep "$match_string" | cut -d"$TAB" -f1) if [ -b "$DEVICE" ]; then blockdevice=yes fi ;; esac } # Lists all filesystems potentially mounted under a given path, # excluding the path itself. list_submounts() { list_mounts | grep "${TAB}${1}/" | cut -d"$TAB" -f2 | sort -r } # Lists all bind mounts of a given file system, # excluding the path itself. list_bindmounts() { if is_bind_mount; then # skip bind mount # we should not umount the original file system via a bind mount return fi match_string="${TAB}${1}${TAB}" if list_mounts | grep "$match_string" >/dev/null 2>&1; then mount_disk=$(list_mounts | grep "$match_string" | cut -d"$TAB" -f1) else return fi if [ -b "$mount_disk" ]; then list_mounts | grep "$mount_disk" | grep -v "$match_string" | cut -d"$TAB" -f2 | sort -r fi } # kernels < 2.6.26 can't handle bind remounts bind_kernel_check() { echo "$options" | grep -w ro >/dev/null 2>&1 || return uname -r | awk -F. ' $1==2 && $2==6 { sub("[^0-9].*","",$3); if ($3<26) exit(1); }' [ $? -ne 0 ] && ocf_log warn "kernel $(uname -r) cannot handle read only bind mounts" } bind_root_mount_check() { if [ "$(df -P "$1" | awk 'END{print $6}')" = "/" ]; then return 1 else return 0 fi } bind_mount() { if is_bind_mount && [ "$options" != "-o bind" ] then bind_kernel_check bind_opts=$(echo "$options" | sed 's/bind/remount/') $MOUNT $bind_opts "$MOUNTPOINT" else true # make sure to return OK fi } is_option() { echo "$OCF_RESKEY_options" | grep -w "$1" >/dev/null 2>&1 } is_fsck_needed() { case $OCF_RESKEY_run_fsck in force) true;; no) false;; ""|auto) case "$FSTYPE" in ext4|ext4dev|ext3|reiserfs|reiser4|nss|xfs|jfs|vfat|fat|nfs4|nfs|aznfs|efs|cifs|smbfs|ocfs2|gfs2|none|lustre|glusterfs|ceph|tmpfs|overlay|overlayfs|rozofs|zfs|cvfs) false;; *) true;; esac;; *) ocf_log warn "Invalid parameter value for fsck: '$OCF_RESKEY_run_fsck'; setting to 'auto'" OCF_RESKEY_run_fsck="auto" is_fsck_needed;; esac } fstype_supported() { local support="$FSTYPE" local rc if [ "X${HOSTOS}" = "XOpenBSD" ];then # skip checking /proc/filesystems for obsd return $OCF_SUCCESS fi if [ -z "$FSTYPE" -o "$FSTYPE" = none ]; then : No FSTYPE specified, rely on the system has the right file-system support already return $OCF_SUCCESS fi # support fuse-filesystems (e.g. GlusterFS) and Amazon Elastic File # System (EFS) case "$FSTYPE" in fuse.*|glusterfs|rozofs) support="fuse";; efs) check_binary "mount.efs"; support="nfs4";; aznfs) check_binary "mount.aznfs"; support="nfs4";; esac if [ "$support" != "$FSTYPE" ]; then ocf_log info "Checking support for $FSTYPE as \"$support\"" fi grep -w "$support"'$' /proc/filesystems >/dev/null if [ $? -eq 0 ]; then # found the fs type return $OCF_SUCCESS fi # if here, we should attempt to load the module and then # check the if the filesystem support exists again. $MODPROBE $support >/dev/null if [ $? -ne 0 ]; then ocf_exit_reason "Couldn't find filesystem $support in /proc/filesystems and failed to load kernel module" return $OCF_ERR_INSTALLED fi # It is possible for the module to load and not be complete initialized # before we check /proc/filesystems again. Give this a few trys before # giving up entirely. for try in $(seq 5); do grep -w "$support"'$' /proc/filesystems >/dev/null if [ $? -eq 0 ] ; then # yes. found the filesystem after doing the modprobe return $OCF_SUCCESS fi ocf_log debug "Unable to find support for $support in /proc/filesystems after modprobe, trying again" sleep 1 done ocf_exit_reason "Couldn't find filesystem $support in /proc/filesystems" return $OCF_ERR_INSTALLED } # # In the case a fresh filesystem is just created from another # node on the shared storage, and is not visible yet. Then try # partprobe to refresh /dev/disk/by-{label,uuid}/* up to date. # # DEVICE can be /dev/xxx, -U, -L # trigger_udev_rules_if_needed() { local refresh_flag="no" local tmp local timeout if [ $blockdevice = "yes" ]; then tmp="$DEVICE" if [ "$DEVICE" != "/dev/null" -a ! -b "$DEVICE" ] ; then refresh_flag="yes" fi else tmp="$(echo $DEVICE|awk '{$1=""; print substr($0,2)}')" case "$DEVICE" in -U*|--uuid*) tmp="/dev/disk/by-uuid/$tmp" ;; -L*|--label*) tmp="/dev/disk/by-label/$tmp" ;; *) # bind mount? return ;; esac [ ! -b "$tmp" ] && refresh_flag="yes" fi [ "$refresh_flag" = "no" ] && return have_binary partprobe && partprobe >/dev/null 2>&1 timeout=${OCF_RESKEY_CRM_meta_timeout:="60000"} timeout=$((timeout/1000)) have_binary udevadm && udevadm settle -t $timeout --exit-if-exists=$tmp return $? } # # START: Start up the filesystem # Filesystem_start() { # Check if there are any mounts mounted under the mountpoint match_string="${TAB}${CANONICALIZED_MOUNTPOINT}" if list_mounts | grep -E "$match_string/\w+" >/dev/null 2>&1; then ocf_log err "There is one or more mounts mounted under $MOUNTPOINT." return $OCF_ERR_CONFIGURED fi # See if the device is already mounted. Filesystem_status case "$?" in $OCF_SUCCESS) ocf_log info "Filesystem $MOUNTPOINT is already mounted." return $OCF_SUCCESS ;; $OCF_ERR_CONFIGURED) return $OCF_ERR_CONFIGURED ;; esac fstype_supported || exit $OCF_ERR_INSTALLED # Check the filesystem & auto repair. # NOTE: Some filesystem types don't need this step... Please modify # accordingly trigger_udev_rules_if_needed if [ $blockdevice = "yes" ]; then if [ "$DEVICE" != "/dev/null" -a ! -b "$DEVICE" ] ; then ocf_exit_reason "Couldn't find device [$DEVICE]. Expected /dev/??? to exist" exit $OCF_ERR_INSTALLED fi if is_fsck_needed; then ocf_log info "Starting filesystem check on $DEVICE" if [ -z "$FSTYPE" ]; then $FSCK -p "$DEVICE" else $FSCK -t "$FSTYPE" -p "$DEVICE" fi # NOTE: if any errors at all are detected, it returns non-zero # if the error is >= 4 then there is a big problem if [ $? -ge 4 ]; then ocf_exit_reason "Couldn't successfully fsck filesystem for $DEVICE" return $OCF_ERR_GENERIC fi fi fi [ -d "$MOUNTPOINT" ] || ocf_run mkdir -p "$MOUNTPOINT" if [ ! -d "$MOUNTPOINT" ] ; then ocf_exit_reason "Couldn't find directory [$MOUNTPOINT] to use as a mount point" exit $OCF_ERR_INSTALLED fi flushbufs "$DEVICE" # Mount the filesystem. case "$FSTYPE" in none) $MOUNT $options $device_opt "$DEVICE" "$MOUNTPOINT" && bind_mount ;; "") $MOUNT $options $device_opt "$DEVICE" "$MOUNTPOINT" ;; *) $MOUNT -t "$FSTYPE" $options $device_opt "$DEVICE" "$MOUNTPOINT" ;; esac if [ $? -ne 0 ]; then ocf_exit_reason "Couldn't mount device [$DEVICE] as $MOUNTPOINT" return $OCF_ERR_GENERIC fi return $OCF_SUCCESS } # end of Filesystem_start get_pids() { local dir=$1 local procs local mmap_procs if is_bind_mount && ocf_is_true "$FORCE_UNMOUNT" && ! bind_root_mount_check "$DEVICE"; then ocf_log debug "Change force_umount from '$FORCE_UNMOUNT' to 'safe'" FORCE_UNMOUNT=safe fi if ocf_is_true "$FORCE_UNMOUNT"; then if [ "X${HOSTOS}" = "XOpenBSD" ]; then fstat | grep $dir | awk '{print $3}' elif [ "X${HOSTOS}" = "XFreeBSD" ]; then $FUSER -c $dir 2>/dev/null else $FUSER -Mm $dir 2>/dev/null fi elif [ "$FORCE_UNMOUNT" = "safe" ]; then # Yes, in theory, ${dir} could contain "intersting" characters # and would need to be quoted for glob (find) and regex (grep). # Don't do that, then. # Avoid /proc/[0-9]*, it may cause "Argument list too long". # There are several ways to filter for /proc/ # -mindepth 1 -not -path "/proc/[0-9]*" -prune -o ... # -path "/proc/[!0-9]*" -prune -o ... # -path "/proc/[0-9]*" -a ... - # the latter seemd to be significantly faster for this one in my naive test. - procs=$(exec 2>/dev/null; - find /proc -path "/proc/[0-9]*" -type l \( -lname "${dir}/*" -o -lname "${dir}" \) -print | - awk -F/ '{print $3}' | uniq) - - # This finds both /proc//maps and /proc//task//maps; - # if you don't want the latter, add -maxdepth. - mmap_procs=$(exec 2>/dev/null; - find /proc -path "/proc/[0-9]*/maps" -print | - xargs -r grep -l " ${dir}/" | awk -F/ '{print $3}' | uniq) - printf "${procs}\n${mmap_procs}" | sort -u + + # find can detect relevant symlinks directly using -lname, + # but we need to grep the content of maps files. + # With many processes, scanning proc may be costly already. + # Do that only once, feed the "maps" files to grep via fifo and + # xargs, and output the matching symlinks directly. + local D + D=$(mktemp -d "$HA_RSCTMP/Filesystem.get_pids.safe.XXXXXX") + mkfifo "$D/maps" + ( + < "$D/maps" xargs -r grep -l " $dir/" | cut -d/ -f3 | uniq & + exec 3> "$D/maps"; # reference for find via /dev/fd/3 + # both xargs and find now have an open fd, + # we can remove the tmpdir and fifo early + rm -rf "$D" + find /proc -path '/proc/[!0-9]*' -prune -o \ + -type f -name maps -fprint /dev/fd/3 -o \ + -type l \( -lname "$dir/*" -o -lname "$dir" \) -print | + cut -d/ -f3 | uniq + ) 2>/dev/null | sort -u fi } signal_processes() { local dir=$1 local sig=$2 local pids pid # fuser returns a non-zero return code if none of the # specified files is accessed or in case of a fatal # error. pids=$(get_pids "$dir") if [ -z "$pids" ]; then ocf_log info "No processes on $dir were signalled. force_unmount is set to '$FORCE_UNMOUNT'" return 1 fi for pid in $pids; do ocf_log info "sending signal $sig to: $(ps -f $pid | tail -1)" kill -s $sig $pid done return 0 } try_umount() { local force_arg="$1" SUB="$2" $UMOUNT $force_arg "$SUB" list_mounts | grep "${TAB}${SUB}${TAB}" >/dev/null 2>&1 || { ocf_log info "unmounted $SUB successfully" return $OCF_SUCCESS } return $OCF_ERR_GENERIC } timeout_child() { local pid="$1" timeout="$2" killer ret # start job in the background that will KILL the given process after timeout expires sleep $timeout && kill -s KILL $pid & killer=$! # block until the child process either exits on its own or gets killed by the above killer pipeline wait $pid ret=$? # ret would be 127 + child exit code if the timeout expired (see "man wait" for more info) [ $ret -lt 128 ] && kill -s KILL $killer return $ret } fs_stop_loop() { local force_arg="$1" SUB="$2" signals="$3" sig send_signal while true; do send_signal=false for sig in $signals; do signal_processes "$SUB" $sig && send_signal=true done $send_signal && sleep $OCF_RESKEY_signal_delay try_umount "$force_arg" "$SUB" && return $OCF_SUCCESS done } fs_stop() { local SUB="$1" timeout=$2 grace_time ret grace_time=$((timeout/2)) # Just walking /proc may take "a long time", even if we don't find any users of this FS. # If dependencies are properly configured, umount should just work. # Only if that fails, try to find and kill processes that still use it. try_umount "" "$SUB" && return $OCF_SUCCESS # try gracefully terminating processes for up to half of the configured timeout fs_stop_loop "" "$SUB" "$OCF_RESKEY_term_signals" & timeout_child $! $grace_time ret=$? [ $ret -eq $OCF_SUCCESS ] && return $ret # try killing them for the rest of the timeout fs_stop_loop "$umount_force" "$SUB" "$OCF_RESKEY_kill_signals" & timeout_child $! $grace_time ret=$? [ $ret -eq $OCF_SUCCESS ] && return $ret # timeout expired ocf_exit_reason "Couldn't unmount $SUB within given timeout" return $OCF_ERR_GENERIC } # # STOP: Unmount the filesystem # Filesystem_stop() { # See if the device is currently mounted Filesystem_status >/dev/null 2>&1 if [ $? -eq $OCF_NOT_RUNNING ]; then # Already unmounted, wonderful. rc=$OCF_SUCCESS else # Wipe the status file, but continue with a warning if # removal fails -- the file system might be read only if [ $OCF_CHECK_LEVEL -eq 20 ]; then rm -f "${STATUSFILE}" if [ $? -ne 0 ]; then ocf_log warn "Failed to remove status file ${STATUSFILE}." fi fi # Determine the real blockdevice this is mounted on (if # possible) prior to unmounting. determine_blockdevice # For networked filesystems, there's merit in trying -f: case "$FSTYPE" in nfs4|nfs|aznfs|efs|cifs|smbfs) umount_force="-f" ;; esac # Umount all sub-filesystems mounted under $MOUNTPOINT/ too. local timeout while read SUB; do ocf_log info "Trying to unmount $SUB" if ocf_is_true "$FAST_STOP"; then timeout=6 else timeout=${OCF_RESKEY_CRM_meta_timeout:="20000"} timeout=$((timeout/1000)) fi fs_stop "$SUB" $timeout rc=$? if [ $rc -ne $OCF_SUCCESS ]; then ocf_exit_reason "Couldn't unmount $SUB, giving up!" fi done <<-EOF $(list_submounts "$CANONICALIZED_MOUNTPOINT"; \ list_bindmounts "$CANONICALIZED_MOUNTPOINT"; \ echo $CANONICALIZED_MOUNTPOINT) EOF fi flushbufs "$DEVICE" return $rc } # end of Filesystem_stop # # STATUS: is the filesystem mounted or not? # Filesystem_status() { local match_string="${TAB}${CANONICALIZED_MOUNTPOINT}${TAB}" local mounted_device=$(list_mounts | grep "$match_string" | awk '{print $1}') if [ -n "$mounted_device" ]; then if [ "X$blockdevice" = "Xyes" ]; then if [ -e "$DEVICE" ] ; then local canonicalized_device="$(readlink -f "$DEVICE")" if [ $? -ne 0 ]; then ocf_exit_reason "Could not canonicalize $DEVICE because readlink failed" exit $OCF_ERR_GENERIC fi else local canonicalized_device="$DEVICE" fi if [ -e "$mounted_device" ] ; then local canonicalized_mounted_device="$(readlink -f "$mounted_device")" if [ $? -ne 0 ]; then ocf_exit_reason "Could not canonicalize $mounted_device because readlink failed" exit $OCF_ERR_GENERIC fi else local canonicalized_mounted_device="$mounted_device" fi if [ "$canonicalized_device" != "$canonicalized_mounted_device" ]; then if ocf_is_probe || [ "$__OCF_ACTION" = "stop" ]; then ocf_log debug "Another device ($mounted_device) is already mounted on $MOUNTPOINT" rc=$OCF_NOT_RUNNING else ocf_exit_reason "Another device ($mounted_device) is already mounted on $MOUNTPOINT" rc=$OCF_ERR_CONFIGURED fi fi else rc=$OCF_SUCCESS msg="$MOUNTPOINT is mounted (running)" fi else rc=$OCF_NOT_RUNNING msg="$MOUNTPOINT is unmounted (stopped)" fi # Special case "monitor" to check whether the UUID cached and # on-disk still match? case "$OP" in status) ocf_log info "$msg";; esac return $rc } # end of Filesystem_status # Note: the read/write tests below will stall in case the # underlying block device (or in the case of a NAS mount, the # NAS server) has gone away. In that case, if I/O does not # return to normal in time, the operation hits its timeout # and it is up to the CRM to initiate appropriate recovery # actions (such as fencing the node). # # MONITOR 10: read the device # Filesystem_monitor_10() { if [ "$blockdevice" = "no" ] ; then ocf_log warn "$DEVICE is not a block device, monitor 10 is noop" return $OCF_SUCCESS fi dd_opts="iflag=direct bs=4k count=1" err_output=$(dd if="$DEVICE" $dd_opts 2>&1 >/dev/null) if [ $? -ne 0 ]; then ocf_exit_reason "Failed to read device $DEVICE" ocf_log err "dd said: $err_output" return $OCF_ERR_GENERIC fi return $OCF_SUCCESS } # # MONITOR 20: write and read a status file # Filesystem_monitor_20() { if [ "$blockdevice" = "no" ] ; then # O_DIRECT not supported on cifs/smbfs dd_opts="oflag=sync bs=4k conv=fsync,sync" else # Writing to the device in O_DIRECT mode is imperative # to bypass caches. dd_opts="oflag=direct,sync bs=4k conv=fsync,sync" fi status_dir=$(dirname "$STATUSFILE") [ -d "$status_dir" ] || mkdir -p "$status_dir" err_output=$(echo "${OCF_RESOURCE_INSTANCE}" | dd of="${STATUSFILE}" $dd_opts 2>&1) if [ $? -ne 0 ]; then ocf_exit_reason "Failed to write status file ${STATUSFILE}" ocf_log err "dd said: $err_output" return $OCF_ERR_GENERIC fi test -f "${STATUSFILE}" if [ $? -ne 0 ]; then ocf_exit_reason "Cannot stat the status file ${STATUSFILE}" return $OCF_ERR_GENERIC fi cat "${STATUSFILE}" > /dev/null if [ $? -ne 0 ]; then ocf_exit_reason "Cannot read the status file ${STATUSFILE}" return $OCF_ERR_GENERIC fi return $OCF_SUCCESS } Filesystem_monitor() { Filesystem_status rc=$? if [ $rc -ne $OCF_SUCCESS ]; then return $rc fi if [ $rc -eq $OCF_SUCCESS -a $OCF_CHECK_LEVEL -gt 0 ]; then case "$OCF_CHECK_LEVEL" in 10) Filesystem_monitor_10; rc=$?;; 20) Filesystem_monitor_20; rc=$?;; *) ocf_exit_reason "unsupported monitor level $OCF_CHECK_LEVEL" rc=$OCF_ERR_CONFIGURED ;; esac fi return $rc } # end of Filesystem_monitor # # VALIDATE_ALL: Are the instance parameters valid? # FIXME!! The only part that's useful is the return code. # This code always returns $OCF_SUCCESS (!) # FIXME!! Needs some tuning to match fstype_supported() (e.g., for # fuse). Can we just call fstype_supported() with a flag like # "no_modprobe" instead? # Filesystem_validate_all() { # Check if the $FSTYPE is workable # NOTE: Without inserting the $FSTYPE module, this step may be imprecise # TODO: This is Linux specific crap. if [ ! -z "$FSTYPE" -a "$FSTYPE" != none ]; then cut -f2 /proc/filesystems | grep "^${FSTYPE}$" >/dev/null 2>&1 if [ $? -ne 0 ]; then modpath=/lib/modules/$(uname -r) moddep=$modpath/modules.dep # Do we have $FSTYPE in modules.dep? cut -d' ' -f1 $moddep \ | grep "^${modpath}.*${FSTYPE}\.k\?o:$" >/dev/null 2>&1 if [ $? -ne 0 ]; then ocf_log info "It seems we do not have $FSTYPE support" fi fi fi # If we are supposed to do monitoring with status files, then # we need a utility to write in O_DIRECT mode. if [ $OCF_CHECK_LEVEL -gt 0 ]; then check_binary dd # Note: really old coreutils version do not support # the "oflag" option for dd. We don't check for that # here. In case dd does not support oflag, monitor is # bound to fail, with dd spewing an error message to # the logs. On such systems, we must do without status # file monitoring. fi #TODO: How to check the $options ? return $OCF_SUCCESS } # # set the blockdevice variable to "no" or "yes" # set_blockdevice_var() { blockdevice=no # these are definitely not block devices case "$FSTYPE" in nfs4|nfs|aznfs|efs|smbfs|cifs|none|glusterfs|ceph|tmpfs|overlay|overlayfs|rozofs|zfs|cvfs|lustre) return;; esac if $(is_option "loop"); then return fi case "$DEVICE" in --uuid=*|--uuid\ *|--label=*|--label\ *) device_opt=$(echo $DEVICE | sed "s/\([[:blank:]]\|=\).*//") DEVICE=$(echo $DEVICE | sed -E "s/$device_opt([[:blank:]]*|=)//") ;; -U*|-L*) # short versions of --uuid/--label device_opt=$(echo $DEVICE | cut -c1-2) DEVICE=$(echo $DEVICE | sed "s/$device_opt[[:blank:]]*//") ;; /dev/null) # Special case for BSC blockdevice=yes ;; *) if [ ! -b "$DEVICE" -a ! -d "$DEVICE" -a "X$OP" != Xstart ] && ! ocf_is_probe ; then ocf_log warn "Couldn't find device [$DEVICE]. Expected /dev/??? to exist" fi if [ ! -d "$DEVICE" ]; then blockdevice=yes fi ;; esac } # Check the arguments passed to this script if [ $# -ne 1 ]; then usage exit $OCF_ERR_ARGS fi # Check the OCF_RESKEY_ environment variables... FORCE_UNMOUNT="yes" if [ -n "${OCF_RESKEY_force_unmount}" ]; then FORCE_UNMOUNT=$OCF_RESKEY_force_unmount fi DEVICE="$OCF_RESKEY_device" case "$DEVICE" in [[:space:]]*|*[[:space:]]) [ "$__OCF_ACTION" = "stop" ] && exit $OCF_SUCCESS ocf_exit_reason "device parameter does not accept leading or trailing whitespace characters" exit $OCF_ERR_CONFIGURED ;; esac FSTYPE=$OCF_RESKEY_fstype if [ ! -z "$OCF_RESKEY_options" ]; then options="-o $OCF_RESKEY_options" fi FAST_STOP=${OCF_RESKEY_fast_stop:="yes"} OP=$1 # These operations do not require instance parameters case $OP in meta-data) meta_data exit $OCF_SUCCESS ;; usage) usage exit $OCF_SUCCESS ;; esac if [ x = x"$DEVICE" ]; then ocf_exit_reason "Please set OCF_RESKEY_device to the device to be managed" exit $OCF_ERR_CONFIGURED fi set_blockdevice_var # Normalize instance parameters: # It is possible that OCF_RESKEY_directory has one or even multiple trailing "/". # But the output of `mount` and /proc/mounts do not. if [ -z "$OCF_RESKEY_directory" ]; then if [ X$OP = "Xstart" -o $blockdevice = "no" ]; then ocf_exit_reason "Please specify the directory" exit $OCF_ERR_CONFIGURED fi else MOUNTPOINT="$(echo "$OCF_RESKEY_directory" | sed 's/\/*$//')" : ${MOUNTPOINT:=/} case "$MOUNTPOINT" in [[:space:]]*|*[[:space:]]) [ "$__OCF_ACTION" = "stop" ] && exit $OCF_SUCCESS ocf_exit_reason "directory parameter does not accept leading or trailing whitespace characters" exit $OCF_ERR_CONFIGURED ;; esac if [ -e "$MOUNTPOINT" ] ; then CANONICALIZED_MOUNTPOINT="$(readlink -f "$MOUNTPOINT")" if [ $? -ne 0 ]; then ocf_exit_reason "Could not canonicalize $MOUNTPOINT because readlink failed" exit $OCF_ERR_GENERIC fi else CANONICALIZED_MOUNTPOINT="$MOUNTPOINT" fi if echo "$CANONICALIZED_MOUNTPOINT" | grep -q "^\s*/\s*$"; then if ocf_is_probe; then ocf_log debug "/ cannot be managed in a cluster" exit $OCF_NOT_RUNNING elif [ "$__OCF_ACTION" = "start" ] || [ "$__OCF_ACTION" = "monitor" ] || [ "$__OCF_ACTION" = "status" ]; then ocf_exit_reason "/ cannot be managed in a cluster" exit $OCF_ERR_CONFIGURED elif [ "$__OCF_ACTION" = "stop" ]; then exit $OCF_SUCCESS fi fi fi # Check to make sure the utilites are found if [ "X${HOSTOS}" != "XOpenBSD" ];then check_binary $MODPROBE check_binary $FUSER fi check_binary $FSCK check_binary $MOUNT check_binary $UMOUNT if [ "$OP" != "monitor" ]; then ocf_log info "Running $OP for $DEVICE on $MOUNTPOINT" fi case $OP in status) Filesystem_status exit $? ;; monitor) Filesystem_monitor exit $? ;; validate-all) Filesystem_validate_all exit $? ;; stop) Filesystem_stop exit $? ;; esac CLUSTERSAFE=0 is_option "ro" && CLUSTERSAFE=2 case "$FSTYPE" in nfs4|nfs|aznfs|efs|smbfs|cifs|none|gfs2|glusterfs|ceph|ocfs2|overlay|overlayfs|tmpfs|cvfs|lustre) CLUSTERSAFE=1 # this is kind of safe too systemd_drop_in "99-Filesystem-remote" "After" "remote-fs.target" ;; # add here CLUSTERSAFE=0 for all filesystems which are not # cluster aware and which, even if when mounted read-only, # could still modify parts of it such as journal/metadata ext4|ext4dev|ext3|reiserfs|reiser4|xfs|jfs) if ocf_is_true "$OCF_RESKEY_force_clones"; then CLUSTERSAFE=2 systemd_drop_in "99-Filesystem-remote" "After" "remote-fs.target" else CLUSTERSAFE=0 # these are not allowed fi ;; esac if ocf_is_clone; then case $CLUSTERSAFE in 0) ocf_exit_reason "DANGER! $FSTYPE on $DEVICE is NOT cluster-aware!" ocf_log err "DO NOT RUN IT AS A CLONE!" ocf_log err "Politely refusing to proceed to avoid data corruption." exit $OCF_ERR_CONFIGURED ;; 2) ocf_log warn "$FSTYPE on $DEVICE is NOT cluster-aware!" if ocf_is_true "$OCF_RESKEY_force_clones"; then ocf_log warn "But we'll let it run because we trust _YOU_ verified it's safe to do so." else ocf_log warn "But we'll let it run because it is mounted read-only." ocf_log warn "Please make sure that it's meta data is read-only too!" fi ;; esac fi case $OP in start) Filesystem_start ;; *) usage exit $OCF_ERR_UNIMPLEMENTED ;; esac exit $?