diff --git a/heartbeat/podman b/heartbeat/podman index 53867bff2..1e3bee482 100755 --- a/heartbeat/podman +++ b/heartbeat/podman @@ -1,628 +1,629 @@ #!/bin/sh # # The podman HA resource agent creates and launches a podman container # based off a supplied podman image. Containers managed by this agent # are both created and removed upon the agent's start and stop actions. # # Copyright (c) 2014 David Vossel # Michele Baldessari # 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_reuse_default="0" : ${OCF_RESKEY_reuse=${OCF_RESKEY_reuse_default}} ####################################################################### meta_data() { cat < 1.0 The podman HA resource agent creates and launches a podman container based off a supplied podman image. Containers managed by this agent are both created and removed upon the agent's start and stop actions. Podman container resource agent. The podman image to base this container off of. podman image The name to give the created container. By default this will be that resource's instance name. podman container name Allow the image to be pulled from the configured podman registry when the image does not exist locally. NOTE, this can drastically increase the time required to start the container if the image repository is pulled over the network. Allow pulling non-local images Add options to be appended to the 'podman run' command which is used when creating the container during the start action. This option allows users to do things such as setting a custom entry point and injecting environment variables into the newly created container. Note the '-d' option is supplied regardless of this value to force containers to run in the background. NOTE: Do not explicitly specify the --name argument in the run_opts. This agent will set --name using either the resource's instance or the name provided in the 'name' argument of this agent. run options Specify a command to launch within the container once it has initialized. run command A comma separated list of directories that the container is expecting to use. The agent will ensure they exist by running 'mkdir -p' Required mount points Specify the full path of a command to launch within the container to check the health of the container. This command must return 0 to indicate that the container is healthy. A non-zero return code will indicate that the container has failed and should be recovered. Note: Using this method for monitoring processes inside a container is not recommended, as containerd tries to track processes running inside the container and does not deal well with many short-lived processes being spawned. Ensure that your container monitors its own processes and terminates on fatal error rather than invoking a command from the outside. monitor command Kill a container immediately rather than waiting for it to gracefully shutdown force kill Allow the container to be reused once it is stopped. By default, containers get removed once they are stopped. Enable this option to have the particular one persist when this happens. reuse container Use transient drop-in files to add extra dependencies to the systemd scopes associated to the container. During reboot, this prevents systemd to stop the container before pacemaker. drop-in dependency END } ####################################################################### REQUIRE_IMAGE_PULL=0 podman_usage() { cat <&1) rc=$? # 125: no container with name or ID ${CONTAINER} found # 126: container state improper (not running) # 127: any other error # 255: podman 2+: container not running case "$rc" in 125|126|255) + ocf_log info "monitor cmd failed (rc=$rc), output: $out, assume container is not running" rc=$OCF_NOT_RUNNING ;; 0) ocf_log debug "monitor cmd passed: exit code = $rc" ;; *) ocf_exit_reason "monitor cmd failed (rc=$rc), output: $out" rc=$OCF_ERR_GENERIC ;; esac return $rc } container_exists() { local rc local out out=$(podman exec ${CONTAINER} $OCF_RESKEY_monitor_cmd 2>&1) rc=$? # 125: no container with name or ID ${CONTAINER} found if [ $rc -ne 125 ]; then return 0 fi return 1 } remove_container() { local rc local execids if ocf_is_true "$OCF_RESKEY_reuse"; then # never remove the container if we have reuse enabled. return 0 fi container_exists if [ $? -ne 0 ]; then # don't attempt to remove a container that doesn't exist return 0 fi ocf_log notice "Cleaning up inactive container, ${CONTAINER}." ocf_run podman rm -v $CONTAINER rc=$? if [ $rc -ne 0 ]; then # due to a podman bug (rhbz#1841485), sometimes a stopped # container can still be associated with Exec sessions, in # which case the "podman rm" has to be forced execids=$(podman inspect $CONTAINER --format '{{len .ExecIDs}}') if [ "$execids" -ne "0" ]; then ocf_log warn "Inactive container ${CONTAINER} has lingering exec sessions. Force-remove it." ocf_run podman rm -f $CONTAINER rc=$? fi fi return $rc } podman_simple_status() { local rc # simple status is implemented via podman exec # everything besides success is considered "not running" monitor_cmd_exec rc=$? if [ $rc -ne $OCF_SUCCESS ]; then rc=$OCF_NOT_RUNNING; fi return $rc } podman_monitor() { # We rely on running podman exec to monitor the container # state because that command seems to be less prone to # performance issue under IO load. # # For probes to work, we expect cmd_exec to be able to report # when a container is not running. Here, we're not interested # in distinguishing whether it's stopped or non existing # (there's function container_exists for that) monitor_cmd_exec return $? } podman_create_mounts() { oldIFS="$IFS" IFS="," for directory in $OCF_RESKEY_mount_points; do mkdir -p "$directory" done IFS="$oldIFS" } podman_container_id() { # Retrieve the container ID by doing a "podman ps" rather than # a "podman inspect", because the latter has performance issues # under IO load. # We could have run "podman start $CONTAINER" to get the ID back # but if the container is stopped, the command will return a # name instead of a container ID. This would break us. podman ps --no-trunc --format '{{.ID}} {{.Names}}' | grep -F -w -m1 "$CONTAINER" | cut -d' ' -f1 } create_transient_drop_in_dependency() { local cid=$1 local rc=$OCF_SUCCESS if [ -z "$cid" ]; then ocf_exit_reason "Container ID not found for \"$CONTAINER\". Not creating drop-in dependency" return $OCF_ERR_GENERIC fi ocf_log info "Creating drop-in dependency for \"$CONTAINER\" ($cid)" for scope in "libpod-$cid.scope.d" "libpod-conmon-$cid.scope.d"; do if [ $rc -eq $OCF_SUCCESS ] && [ ! -d /run/systemd/transient/"$scope" ]; then mkdir -p /run/systemd/transient/"$scope" && \ printf "[Unit]\nBefore=pacemaker.service" > /run/systemd/transient/"$scope"/dep.conf && \ chmod ago+r /run/systemd/transient/"$scope" /run/systemd/transient/"$scope"/dep.conf rc=$? fi done if [ $rc -ne $OCF_SUCCESS ]; then ocf_log error "Could not create drop-in dependency for \"$CONTAINER\" ($cid)" else systemctl daemon-reload rc=$? if [ $rc -ne $OCF_SUCCESS ]; then ocf_log error "Could not refresh service definition after creating drop-in for \"$CONTAINER\"" fi fi return $rc } run_new_container() { local opts=$1 local image=$2 local cmd=$3 local rc ocf_log info "running container $CONTAINER for the first time" out=$(podman run $opts $image $cmd 2>&1) rc=$? if [ -n "$out" ]; then out="$(echo "$out" | tr -s ' \t\r\n' ' ')" if [ $rc -eq 0 ]; then ocf_log info "$out" else ocf_log err "$out" fi fi if [ $rc -eq 125 ]; then # If an internal podman error occurred, it might be because # the internal storage layer still references an old container # with the same name, even though podman itself thinks there # is no such container. If so, purge the storage layer to try # to clean the corruption and try again. if echo "$out" | grep -q "unknown.*flag"; then ocf_exit_reason "$out" return $rc fi ocf_log warn "Internal podman error while creating new container $CONTAINER. Retrying." ocf_run podman rm --storage $CONTAINER ocf_run podman run $opts $image $cmd rc=$? elif [ $rc -eq 127 ]; then # rhbz#1972209: podman 3.0.x seems to be hit by a race # where the cgroup is not yet set up properly when the OCI # runtime configures the container. If that happens, recreate # the container as long as we get the same error code or # until start timeout preempts us. while [ $rc -eq 127 ] && (echo "$out" | grep -q "cgroup.*scope not found") ; do ocf_log warn "Internal podman error while assigning cgroup. Retrying." # Arbitrary sleep to prevent consuming all CPU while looping sleep 1 podman rm -f "$CONTAINER" out=$(podman run $opts $image $cmd 2>&1) rc=$? done # Log the created container ID if it succeeded if [ $rc -eq 0 ]; then ocf_log info "$out" fi fi return $rc } podman_start() { local cid local rc podman_create_mounts local run_opts="-d --name=${CONTAINER}" # check to see if the container has already started podman_simple_status if [ $? -eq $OCF_SUCCESS ]; then return $OCF_SUCCESS fi if [ -n "$OCF_RESKEY_run_opts" ]; then run_opts="$run_opts $OCF_RESKEY_run_opts" fi if [ $REQUIRE_IMAGE_PULL -eq 1 ]; then ocf_log notice "Beginning pull of image, ${OCF_RESKEY_image}" podman pull "${OCF_RESKEY_image}" if [ $? -ne 0 ]; then ocf_exit_reason "failed to pull image ${OCF_RESKEY_image}" return $OCF_ERR_GENERIC fi fi if ocf_is_true "$OCF_RESKEY_reuse" && container_exists; then ocf_log info "starting existing container $CONTAINER." ocf_run podman start $CONTAINER else # make sure any previous container matching our container name is cleaned up first. # we already know at this point it wouldn't be running remove_container run_new_container "$run_opts" $OCF_RESKEY_image "$OCF_RESKEY_run_cmd" if [ $? -eq 125 ]; then return $OCF_ERR_GENERIC fi fi rc=$? # if the container was stopped or didn't exist before, systemd # removed the libpod* scopes. So always try to recreate the drop-ins if [ $rc -eq 0 ] && ocf_is_true "$OCF_RESKEY_drop_in_dependency"; then cid=$(podman_container_id) create_transient_drop_in_dependency "$cid" rc=$? fi if [ $rc -ne 0 ]; then ocf_exit_reason "podman failed to launch container (rc: $rc)" return $OCF_ERR_GENERIC fi # wait for monitor to pass before declaring that the container is started while true; do podman_simple_status if [ $? -ne $OCF_SUCCESS ]; then ocf_exit_reason "Newly created podman container exited after start" return $OCF_ERR_GENERIC fi monitor_cmd_exec if [ $? -eq $OCF_SUCCESS ]; then ocf_log notice "Container $CONTAINER started successfully" return $OCF_SUCCESS fi ocf_exit_reason "waiting on monitor_cmd to pass after start" sleep 1 done } podman_stop() { local timeout=60 local rc podman_simple_status if [ $? -eq $OCF_NOT_RUNNING ]; then remove_container return $OCF_SUCCESS fi if [ -n "$OCF_RESKEY_CRM_meta_timeout" ]; then timeout=$((($OCF_RESKEY_CRM_meta_timeout/1000) -10 )) if [ $timeout -lt 10 ]; then timeout=10 fi fi if ocf_is_true "$OCF_RESKEY_force_kill"; then ocf_run podman kill $CONTAINER rc=$? else ocf_log debug "waiting $timeout second[s] before killing container" ocf_run podman stop -t=$timeout $CONTAINER rc=$? # on stop, systemd will automatically delete any transient # drop-in conf that has been created earlier fi if [ $rc -ne 0 ]; then # If the stop failed, it could be because the controlling conmon # process died unexpectedly. If so, a generic error code is returned # but the associated container exit code is -1. If that's the case, # assume there's no failure and continue with the rm as usual. if [ $rc -eq 125 ] && \ podman inspect --format '{{.State.Status}}:{{.State.ExitCode}}' $CONTAINER | grep -wq "stopped:-1"; then ocf_log warn "Container ${CONTAINER} had an unexpected stop outcome. Trying to remove it anyway." else ocf_exit_reason "Failed to stop container, ${CONTAINER}, based on image, ${OCF_RESKEY_image}." return $OCF_ERR_GENERIC fi fi remove_container if [ $? -ne 0 ]; then ocf_exit_reason "Failed to remove stopped container, ${CONTAINER}, based on image, ${OCF_RESKEY_image}." return $OCF_ERR_GENERIC fi return $OCF_SUCCESS } image_exists() { podman image exists "${OCF_RESKEY_image}" if [ $? -eq 0 ]; then # image found return 0 fi if ocf_is_true "$OCF_RESKEY_allow_pull"; then REQUIRE_IMAGE_PULL=1 ocf_log notice "Image (${OCF_RESKEY_image}) does not exist locally but will be pulled during start" return 0 fi # image not found. return 1 } podman_validate() { check_binary podman if [ -z "$OCF_RESKEY_image" ]; then ocf_exit_reason "'image' option is required" exit $OCF_ERR_CONFIGURED fi image_exists if [ $? -ne 0 ]; then ocf_exit_reason "base image, ${OCF_RESKEY_image}, could not be found." exit $OCF_ERR_CONFIGURED fi return $OCF_SUCCESS } # TODO : # When a user starts plural clones in a node in globally-unique, a user cannot appoint plural name parameters. # When a user appoints reuse, the resource agent cannot connect plural clones with a container. if ocf_is_true "$OCF_RESKEY_CRM_meta_globally_unique"; then if [ -n "$OCF_RESKEY_name" ]; then if [ -n "$OCF_RESKEY_CRM_meta_clone_node_max" ] && [ "$OCF_RESKEY_CRM_meta_clone_node_max" -ne 1 ] then ocf_exit_reason "Cannot make plural clones from the same name parameter." exit $OCF_ERR_CONFIGURED fi if [ -n "$OCF_RESKEY_CRM_meta_master_node_max" ] && [ "$OCF_RESKEY_CRM_meta_master_node_max" -ne 1 ] then ocf_exit_reason "Cannot make plural master from the same name parameter." exit $OCF_ERR_CONFIGURED fi fi : ${OCF_RESKEY_name=`echo ${OCF_RESOURCE_INSTANCE} | tr ':' '-'`} else : ${OCF_RESKEY_name=${OCF_RESOURCE_INSTANCE}} fi CONTAINER=$OCF_RESKEY_name # Note: we currently monitor podman containers by with the "podman exec" # command, so make sure that invocation is always valid by enforcing the # exec command to be non-empty : ${OCF_RESKEY_monitor_cmd:=/bin/true} # When OCF_RESKEY_drop_in_dependency is not populated, we # look at another file-based way of enabling the option. # Otherwise, consider it disabled. if [ -z "$OCF_RESKEY_drop_in_dependency" ]; then if [ -f "/etc/sysconfig/podman_drop_in" ] || \ [ -f "/etc/default/podman_drop_in" ]; then OCF_RESKEY_drop_in_dependency=yes fi fi case $__OCF_ACTION in meta-data) meta_data exit $OCF_SUCCESS;; start) podman_validate podman_start;; stop) podman_stop;; monitor) podman_monitor;; validate-all) podman_validate;; usage|help) podman_usage exit $OCF_SUCCESS ;; *) podman_usage exit $OCF_ERR_UNIMPLEMENTED ;; esac rc=$? ocf_log debug "${OCF_RESOURCE_INSTANCE} $__OCF_ACTION : $rc" exit $rc