#!/bin/bash
CONTAINER_IMAGE=${CONTAINER_IMAGE:-"docker.io/nyang23/haleap:master"}
PROJECT_PATH=$(dirname $(dirname `realpath $0`))
PROJECT_INSIDE="/opt/crmsh"
COROSYNC_CONF="/etc/corosync/corosync.conf"
COROSYNC_AUTH="/etc/corosync/authkey"
HA_NETWORK_FIRST="ha_network_first"
HA_NETWORK_SECOND="ha_network_second"
declare -a HA_NETWORK_ARRAY
declare -a HA_NETWORK_V6_ARRAY
HA_NETWORK_ARRAY[0]=$HA_NETWORK_FIRST
HA_NETWORK_ARRAY[1]=$HA_NETWORK_SECOND
HA_NETWORK_V6_ARRAY[0]="2001:db8:10::/64"
HA_NETWORK_V6_ARRAY[1]="2001:db8:20::/64"
BEHAVE_CASE_DIR="$(dirname $0)/features/"
BEHAVE_CASE_EXCLUDE="sbd|ocfs2|gfs2"

read -r -d '' SSHD_CONFIG_AZURE << EOM
PermitRootLogin no
AuthorizedKeysFile .ssh/authorized_keys
ChallengeResponseAuthentication no
UsePAM yes
X11Forwarding yes
ClientAliveInterval 180
Subsystem sftp	/usr/lib/ssh/sftp-server
AcceptEnv LANG LC_CTYPE LC_NUMERIC LC_TIME LC_COLLATE LC_MONETARY LC_MESSAGES
AcceptEnv LC_PAPER LC_NAME LC_ADDRESS LC_TELEPHONE LC_MEASUREMENT
AcceptEnv LC_IDENTIFICATION LC_ALL
PasswordAuthentication no
EOM

read -r -d '' COROSYNC_CONF_TEMPLATE << EOM
totem {
    version: 2
    cluster_name: cluster1
    transport: knet
    crypto_cipher: aes256
    crypto_hash: sha256
}

nodelist {
}

quorum {
    provider: corosync_votequorum
}

logging {
    to_logfile: yes
    logfile: /var/log/cluster/corosync.log
    to_syslog: yes
    timestamp: on
}
EOM


fatal() {
	error $*
	exit 1
}


error() {
	echo "ERROR: $*"
}


warning() {
	echo "WARNING: $*"
}


info() {
	echo "INFO: $*"
}


is_number() {
	num=$1
	test ! -z "$num" && test "$num" -eq "$num" 2> /dev/null && test "$num" -gt 0 2> /dev/null
}


get_test_case_array() {
	test -d $BEHAVE_CASE_DIR || fatal "Cannot find '$BEHAVE_CASE_DIR'"
	ls $BEHAVE_CASE_DIR|grep "\.feature"|grep -Ev "$BEHAVE_CASE_EXCLUDE"
}


echo_test_cases() {
	case_arry=`get_test_case_array`
	echo "Index|File Name|Description"
	index=1
	 for f in ${case_arry[@]};do
		desc=`awk -F: '/Feature/{print $2}' $BEHAVE_CASE_DIR/$f`
		printf "%3s    %-40s %-60s\n" $index $f "$desc"
		index=$(($index+1))
	done
	printf "%3s    %-40s  %-60s\n" $index "regression test" "Original regression test"
}


usage_and_exit() {
	prog=`basename $0`
	cat <<END
Usage: $prog [OPTIONS]|[TESTCASE INDEX]
$prog is a tool for developers to setup the cluster in containers to run functional tests.
The container image is based on Tumbleweed with preinstalled packages of the cluster stack include pacemaker/corosync/crmsh and many others.
Users can make the code change under crmsh.git including test cases. This tool will pick up the code change and "make install" to all running containers.

OPTIONS:
  -h, --help		Show this help message and exit
  -l            	List existing functional test cases and exit
  -n NUM        	Only setup a cluster with NUM nodes(containers)
  -x			Don't config corosync on containers(with -n option)
  -d			Cleanup the cluster containers
  -u			Create normal users, and Azure like ssh environment
  -q			Create a qnetd node(with -n and -x option)

EXAMPLES:
To launch 2 nodes with the running cluster with the very basic corosync.conf
# crmsh.git/test/run-functional-tests -n 2

To launch 2 nodes without the cluster stack running to play with "crm cluster init/join"
# crmsh.git/run-functional-tests -n 2 -x

To launch 2 nodes without the cluster stack running, and a qnetd node(named 'qnetd-node')
# crmsh.git/run-functional-tests -n 2 -x -q

To list the existing test cases. Users could add his own new test cases.
# crmsh.git/test/run-functional-tests -l

To run a single or a number of functional test cases
# crmsh.git/test/run-functional-tests 1
# crmsh.git/test/run-functional-tests 1 2 3

To clean up the all containers which are generated by this tool
# crmsh.git/test/run-functional-tests -d
END
	exit 1
}


podman_exec() {
	name=$1
	cmd=$2
	podman exec -t $name /bin/sh -c "$cmd"
}

set_sshd_config_like_in_azure() {
	node_name=$1
	podman_exec $node_name "echo \"$SSHD_CONFIG_AZURE\" > /etc/ssh/sshd_config"
	podman_exec $node_name "systemctl restart sshd.service"
}

create_custom_user() {
	user_name=$1
	user_id=$2
	podman_exec $node_name "useradd -m -s /bin/bash ${user_name} 2>/dev/null"
	podman_exec $node_name "echo \"${user_name} ALL=(ALL) NOPASSWD: ALL\" > /etc/sudoers.d/${user_name}"
	podman_exec $node_name "chmod 0440 /etc/sudoers.d/${user_name}"
	podman_exec $node_name "echo 'export PATH=\$PATH:/usr/sbin/' >> ~${user_name}/.bashrc"
	podman_exec $node_name "echo -e \"linux\\nlinux\" | passwd ${user_name} 2>/dev/null"
	podman_exec $node_name "cp -r /root/.ssh ~${user_name}/ && chown ${user_name}:haclient -R ~${user_name}/.ssh"
	info "Create user '$user_name' on $node_name"
}

create_alice_bob_carol() {
	# Custom users are alice, bob and carol and they are as important as the root
	# and eventually they should be already in the docker image
	# However now, let's create them here
	create_custom_user "alice" "1000"
	create_custom_user "bob" "1001"
	create_custom_user "carol" "1002"
}

deploy_ha_node() {
	node_name=$1
	podman_options="--name $node_name -h $node_name --network ha_network_first --systemd always --shm-size 1g"
        # CAP_SYS_NICE for corosync to set realtime priority
        # CAP_AUDIT_CONTROL for sshd
        # CAP_NET_ADMIN for firewall and virtual ip
        podman_capabilties="--cap-add CAP_SYS_ADMIN --cap-add CAP_SYS_NICE --cap-add CAP_AUDIT_CONTROL --cap-add CAP_NET_ADMIN"
        if [ -d /sys/kernel/security/apparmor ]; then
            if [ -f /etc/apparmor.d/podman ]; then
                podman_security="--security-opt=apparmor=podman"
            else
                podman_security="--security-opt=apparmor=unconfined"
            fi
        else
            podman_security=""
        fi
	make_cmd="cd $PROJECT_INSIDE;./autogen.sh && ./configure --prefix /usr && make install && make install-crmconfDATA prefix= && cp /usr/bin/crm /usr/sbin"

	info "Deploying \"$node_name\"..."
	podman run --rm -d $podman_options $podman_capabilties $podman_security $CONTAINER_IMAGE
        podman network connect ha_network_second $node_name

	if [ "$node_name" != "qnetd-node" ];then
		rm_qnetd_cmd="rpm -q corosync-qnetd && rpm -e corosync-qnetd"
		podman_exec $node_name "$rm_qnetd_cmd"
	fi
	podman_exec $node_name "rm -rf /run/nologin"
	podman_exec $node_name "echo 'StrictHostKeyChecking no' >> /etc/ssh/ssh_config"
	podman_exec $node_name "systemctl start sshd.service"

	if [[ "$node_name" != "qnetd-node" && ! "$node_name" =~ ^pcmk-remote-node[0-9]$ ]];then
	  podman cp $PROJECT_PATH $node_name:/opt/crmsh
	  info "Building crmsh on \"$node_name\"..."
	  podman_exec $node_name "$make_cmd" || \
		fatal "Building failed on $node_name!"
	  podman_exec $node_name "chown hacluster:haclient -R /var/log/crmsh"
	  podman_exec $node_name "chmod g+w -R /var/log/crmsh"
	  create_alice_bob_carol
	  if [ "$NORMAL_USER_FLAG" -eq 1 ];then
		set_sshd_config_like_in_azure $node_name
          fi
	  podman_exec $node_name "sed -i 's/; \[report\]/[report]/' /etc/crm/crm.conf"
	  podman_exec $node_name "sed -i 's/; compress_prog = .*$/compress_prog = bzip2/g' /etc/crm/crm.conf"
	else
	  podman_exec $node_name "useradd -m -s /bin/bash alice 2>/dev/null"
	  podman_exec $node_name "echo \"alice ALL=(ALL) NOPASSWD:ALL\" > /etc/sudoers.d/alice"
	  podman_exec $node_name "cp -r /root/.ssh ~alice/ && chown alice:users -R ~alice/.ssh"
	  info "Create user 'alice' on $node_name"
	  [ "$NORMAL_USER_FLAG" -eq 1 ] && set_sshd_config_like_in_azure $node_name
	fi
}


create_node() {
        echo 16777216 > /proc/sys/net/core/rmem_max
        echo 16777216 > /proc/sys/net/core/wmem_max
        info "Create ha specific container networks..."
	for index in ${!HA_NETWORK_ARRAY[@]};do
		network=${HA_NETWORK_ARRAY[$index]}
		podman network create --ipv6 --subnet ${HA_NETWORK_V6_ARRAY[$index]} $network
	done

	for node in $*;do
		deploy_ha_node $node &
	done
	wait
}


config_cluster() {
	node_num=$#
	insert_str=""
	for node in $*;do
		ip=`podman container inspect $node -f "{{.NetworkSettings.Networks.$HA_NETWORK_ARRAY.IPAddress}}"|tr -d "\r"`
		container_ip_array+=($ip)
	done

	for i in $(seq $node_num -1 1);do
		ip=`echo ${container_ip_array[$((i-1))]}|awk -F/ '{print $1}'`
		insert_str+="\\n\\tnode {\n\t\tring0_addr: $ip\n\t\tnodeid: $i\n\t}"
	done
	corosync_conf_str=$(sed "/nodelist/a \\${insert_str}" <(echo "$COROSYNC_CONF_TEMPLATE"))
	if [ $node_num -eq 2 ];then
		corosync_conf_str=$(sed "/corosync_votequorum/a \\\\ttwo_node: 1" <(echo "$corosync_conf_str"))
	fi

	info "Copy corosync.conf to $*"
	for node in $*;do
		if [ $node == $1 ];then
			podman_exec $1 "echo \"$corosync_conf_str\" >> $COROSYNC_CONF"
			podman_exec $1 "corosync-keygen -l -k $COROSYNC_AUTH &> /dev/null"
		else
			while :
			do
				podman_exec $1 "ssh -T -o Batchmode=yes $node true &> /dev/null" && break
				sleep 1
			done
			podman_exec $1 "scp -p $COROSYNC_CONF $COROSYNC_AUTH $node:/etc/corosync &> /dev/null"
		fi
	done
}


start_cluster() {
	for node in $*;do
		podman_exec $node "crm cluster enable && crm cluster start"
		if [ "$?" -eq 0 ];then
			info "Cluster service started on \"$node\""
		else
			fatal "Failed to start cluster service on \"$node\""
		fi
	done
}


container_already_exists() {
	podman ps -a|grep -q "$1"
	if [ "$?" -eq 0 ];then
		fatal "Container \"$1\" already running"
	fi
}


setup_cluster() {
	hanodes_arry=()
	is_number $1
	if [ "$?" -eq 0 ];then
		for i in $(seq 1 $1);do
			hanodes_arry+=("hanode$i")
		done
	else
		hanodes_arry=($*)
	fi

	if [ $WITH_QNETD_NODE -eq 1 ];then
	  create_node ${hanodes_arry[@]} "qnetd-node"
	else
	  create_node ${hanodes_arry[@]}
	fi

	[ "$CONFIG_COROSYNC_FLAG" -eq 0 ] && return
	config_cluster ${hanodes_arry[@]}
	start_cluster ${hanodes_arry[@]}
	podman_exec "hanode1" "crm configure property stonith-enabled=false"
}


cleanup_cluster() {
        info "Cleanup ha specific containers..."
        podman ps --format json | jq -r '.[].Names[]' | grep -E 'hanode.*|qnetd-node|pcmk-remote-node[0-9]' | xargs podman stop
        info "Cleanup ha specific container networks..."
	exist_network_array=()
	for network in ${HA_NETWORK_ARRAY[@]};do
		podman network ls|grep -q $network && exist_network_array+=($network)
	done
	for network in ${exist_network_array[@]};do
		podman network rm $network
	done
}


adjust_test_case() {
	node_name=$1
	replace_arry=(`podman_exec $node_name "grep -o -E '@(hanode[0-9]+|qnetd-node)\.ip6?\.[0-9]' $2|sort -u"`)
	for item in ${replace_arry[@]};do
		item_str=${item##@}
		node=`echo $item_str|cut -d "." -f 1`
		ip_version=`echo $item_str|cut -d "." -f 2|tr -d "\r"`
		ip_search_str="IPAddress"
		if [ "$ip_version" == "ip6" ];then
			ip_search_str="GlobalIPv6Address"
		fi
		index=`echo $item_str|cut -d "." -f 3|tr -d "\r"`
                ip=`podman container inspect $node -f "{{.NetworkSettings.Networks.${HA_NETWORK_ARRAY[$index]}.$ip_search_str}}"|tr -d "\r"`
		item=`echo $item|tr -d "\r"`
		podman_exec $node_name "sed -i s/$item/$ip/g $2"
	done

	vip_replace_array=(`podman_exec $node_name "grep -o -E '@vip\.[0-9]' $2|sort -u|dos2unix"`)
	for item in ${vip_replace_array[@]};do
		index=`echo $item|cut -d "." -f 2|tr -d "\r"`
		suffix=$((123+index))
		ip=`podman container inspect $node_name -f "{{.NetworkSettings.Networks.${HA_NETWORK_ARRAY[0]}.IPAddress}}"|tr -d "\r"`
		vip=`echo $ip|sed "s/\.[0-9][0-9]*$/\.$suffix/g"|tr -d "\r"`
		item=`echo $item|tr -d "\r"`
		podman_exec $node_name "sed -i s/$item/$vip/g $2"
	done
}


run_origin_regression_test() {
	CONFIG_COROSYNC_FLAG=0
	setup_cluster "hanode1"
	podman_exec "hanode1" "echo -e '[core]\nhas_ra_advised_op = yes\nhas_fa_advised_op = yes' > /etc/crm/crm.conf"
	podman_exec "hanode1" "sh /usr/share/crmsh/tests/regression.sh"
	return $?
}


prepare_coverage_env() {
	for node in $*; do
            podman cp "$PROJECT_PATH"/test/features/coverage/sitecustomize.py "$node":/usr/lib/python3.13/site-packages/sitecustomize.py
	done
}


fetch_coverage_report() {
	local unique_id=$(dd if=/dev/urandom count=1 bs=6 | base64 | tr '+/' '-_')
	for node in $*; do
		podman_exec "$node" "coverage combine; coverage xml -o /opt/coverage.xml"
		# see https://github.com/codecov/codecov-cli/blob/master/codecov_cli/services/upload/coverage_file_finder.py
		podman cp "$node":/opt/coverage.xml coverage."$unique_id"."$node".xml
	done
}


WITH_QNETD_NODE=0
NORMAL_USER_FLAG=0
CONFIG_COROSYNC_FLAG=1
SETUP_N_NODES_CLUSTER=0
options=$(getopt -l "help" -o "hldxuqn:" -- "$@")
eval set -- "$options"
while true;do
case $1 in
-h|--help) usage_and_exit;;
-l)
	echo_test_cases
	exit 0
	;;
-d)
	cleanup_cluster
	exit $?
	;;
-x)
	CONFIG_COROSYNC_FLAG=0
	shift
	;;
-u)
	NORMAL_USER_FLAG=1
	shift
	;;
-q)
	WITH_QNETD_NODE=1
	shift
	;;
-n)
	shift
	is_number $1 || fatal "-n option need a number larger than 0"
	SETUP_N_NODES_CLUSTER=$1
	shift
	;;
--)
	shift
	break
	;;
esac
done

if [ $SETUP_N_NODES_CLUSTER -ge 1 ];then
	setup_cluster $SETUP_N_NODES_CLUSTER
	exit $?
fi

if [ "$#" -eq 0 ];then
	usage_and_exit
fi

# used by github action
if [ "$1" == "_get_index_of" ];then
	shift
	pattern=""
	for item in $*;do
		pattern+="${item}|"
	done
	echo_test_cases|grep -E "(${pattern%%|})(\.feature|\s+Original regression test)"|awk '{print $1}'|tr -s '\n' ' '
	exit 0
fi

for case_num in $*;do
	echo_test_cases|grep -E "\s+$case_num\s" &> /dev/null
	if [ "$?" -ne 0 ];then
		error "\"$case_num\" is an invalid index"
		echo_test_cases
		exit 1
	fi
done

for case_num in $*;do
	if [ "$case_num" -ne $1 ];then
		cleanup_cluster
		echo
	fi
	test_case_array=(`get_test_case_array`)
	if [ $case_num -gt ${#test_case_array[*]} ];then
		run_origin_regression_test || exit 1
		continue
	fi
	case_file=$BEHAVE_CASE_DIR/${test_case_array[$((case_num-1))]}
	case_file_in_container="$PROJECT_INSIDE/test/features/`basename $case_file`"
	node_arry=(`awk -F: '/Need nodes/{print $2}' $case_file`)
	CONFIG_COROSYNC_FLAG=0
	setup_cluster ${node_arry[@]}
	adjust_test_case ${node_arry[0]} $case_file_in_container
	echo
        prepare_coverage_env "${node_arry[@]}"
	if [ "$NORMAL_USER_FLAG" -eq 0 ];then
		info "Running \"$case_file_in_container\" under 'root'..."
		podman_exec ${node_arry[0]} "behave --no-logcapture $case_file_in_container || exit 1" || exit 1
	else
		info "Running \"$case_file_in_container\" under normal user 'alice'..."
                podman_exec ${node_arry[0]} "su - alice -c 'sudo behave --no-logcapture $case_file_in_container || exit 1'" || exit 1
	fi
	fetch_coverage_report "${node_arry[@]}"
	echo
done