feat: 自启动与自恢复机制
This commit is contained in:
97
scripts/boot/blitz-fault-inject.sh
Normal file
97
scripts/boot/blitz-fault-inject.sh
Normal file
@@ -0,0 +1,97 @@
|
||||
#!/usr/bin/env bash
|
||||
set -euo pipefail
|
||||
|
||||
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
|
||||
# shellcheck disable=SC1091
|
||||
source "${SCRIPT_DIR}/common.sh"
|
||||
|
||||
STEP="fault-inject"
|
||||
B_SIDE_SERVICE="blitz-b-side-omnid.service"
|
||||
ROS_SERVICE="blitz-ros-receiver.service"
|
||||
|
||||
main_pid_for_service() {
|
||||
local service_name="$1"
|
||||
systemctl show --property MainPID --value "${service_name}"
|
||||
}
|
||||
|
||||
require_running_pid() {
|
||||
local service_name="$1"
|
||||
local pid
|
||||
|
||||
pid="$(main_pid_for_service "${service_name}")"
|
||||
if [[ -z "${pid}" || "${pid}" == "0" ]]; then
|
||||
blitz_log "${STEP}" "lookup-pid" "failure" "service=${service_name}" 1
|
||||
exit 1
|
||||
fi
|
||||
printf '%s\n' "${pid}"
|
||||
}
|
||||
|
||||
write_fault_flag() {
|
||||
local flag_name="$1"
|
||||
local flag_path="${BLITZ_RUNTIME_DIR}/${flag_name}"
|
||||
printf '%s\n' "$(date +%s)" > "${flag_path}"
|
||||
blitz_log "${STEP}" "flag-on" "success" "path=${flag_path}" 0
|
||||
}
|
||||
|
||||
clear_fault_flag() {
|
||||
local flag_name="$1"
|
||||
local flag_path="${BLITZ_RUNTIME_DIR}/${flag_name}"
|
||||
rm -f "${flag_path}"
|
||||
blitz_log "${STEP}" "flag-off" "success" "path=${flag_path}" 0
|
||||
}
|
||||
|
||||
blitz_load_boot_env
|
||||
blitz_require_root "${STEP}"
|
||||
blitz_prepare_runtime_dir
|
||||
|
||||
case "${1:-}" in
|
||||
bside-crash)
|
||||
kill -9 "$(require_running_pid "${B_SIDE_SERVICE}")"
|
||||
;;
|
||||
bside-process-freeze)
|
||||
kill -STOP "$(require_running_pid "${B_SIDE_SERVICE}")"
|
||||
;;
|
||||
bside-video-thread-stall)
|
||||
write_fault_flag "fault-injection-bside-video-thread-stall"
|
||||
;;
|
||||
bside-control-thread-stall)
|
||||
write_fault_flag "fault-injection-bside-control-thread-stall"
|
||||
;;
|
||||
ros-crash)
|
||||
kill -9 "$(require_running_pid "${ROS_SERVICE}")"
|
||||
;;
|
||||
ros-freeze)
|
||||
kill -STOP "$(require_running_pid "${ROS_SERVICE}")"
|
||||
;;
|
||||
network-down)
|
||||
if [[ "${BLITZ_WATCHDOG_ALLOW_FAULT_INJECTION}" != "1" ]]; then
|
||||
blitz_log "${STEP}" "network-down" "failure" "set BLITZ_WATCHDOG_ALLOW_FAULT_INJECTION=1 first" 1
|
||||
exit 1
|
||||
fi
|
||||
case "${2:-}" in
|
||||
on)
|
||||
write_fault_flag "fault-injection-network-down"
|
||||
;;
|
||||
off)
|
||||
clear_fault_flag "fault-injection-network-down"
|
||||
;;
|
||||
*)
|
||||
echo "usage: $0 network-down on|off" >&2
|
||||
exit 2
|
||||
;;
|
||||
esac
|
||||
;;
|
||||
*)
|
||||
cat <<'EOF'
|
||||
usage:
|
||||
blitz-fault-inject.sh bside-crash
|
||||
blitz-fault-inject.sh bside-process-freeze
|
||||
blitz-fault-inject.sh bside-video-thread-stall
|
||||
blitz-fault-inject.sh bside-control-thread-stall
|
||||
blitz-fault-inject.sh ros-crash
|
||||
blitz-fault-inject.sh ros-freeze
|
||||
blitz-fault-inject.sh network-down on|off
|
||||
EOF
|
||||
exit 2
|
||||
;;
|
||||
esac
|
||||
Reference in New Issue
Block a user