fix(dbbackup): 缺 MinecraftServer 导出时重试、CLI 失败退出、面板与 watchdog 告警，异地快照要求完整

deploy/bootstrap.sh

@@ -4074,7 +4074,9 @@ migrate_host_postgres() {
   remember_temp "$fresh"
   write_felis_toml "$legacy" "127.0.0.1:5432"
   write_felis_toml "$fresh" "127.0.0.1:${PG_HOST_PORT}" "${CONTROL_NS}/${PG_DEPLOYMENT}"
-  out="$("$HOST_BIN" db backup -config "$legacy" -dir "$FELIS_DB_BACKUP_DIR" -label pre-pg-move -keep 0)" \
+  # The move carries the database alone; the MinecraftServer objects stay in the
+  # cluster, so a cluster slow to answer their export must not stop it.
+  out="$("$HOST_BIN" db backup -config "$legacy" -dir "$FELIS_DB_BACKUP_DIR" -label pre-pg-move -keep 0 -no-servers)" \
     || die "could not take a bundle of the host database; nothing was moved"
   bundle="$(printf '%s\n' "$out" | sed -n 's/^felis db backup: wrote //p' | tail -n 1)"
   [ -n "$bundle" ] && [ -f "$bundle" ] || die "felis db backup reported no bundle; nothing was moved"

deploy/bootstrap_test.sh

@@ -3534,7 +3534,7 @@ run_move() { # holds(0/1) backup(ok|silent|fail) restore-exit compare-exit
 out="$(run_move 1 ok)"
 calls="$(cat "$mgdir/calls")"
 expect "the bundle is taken from the host server, never pruned" \
-  "FELIS db backup [127.0.0.1:5432 no-deployment] -dir $mgdir/db -label pre-pg-move -keep 0" "$calls"
+  "FELIS db backup [127.0.0.1:5432 no-deployment] -dir $mgdir/db -label pre-pg-move -keep 0 -no-servers" "$calls"
 expect "the bundle is restored into the pod, which served nobody yet" \
   "FELIS db restore [127.0.0.1:15432 felis/felis-postgres] -dir $mgdir/db -yes -no-safety-backup $mgdir/db/b.tar" "$calls"
 before "writers are stopped before the bundle" "QUIESCE" "FELIS db backup" "$calls"