2个provider和Reference用nacos折开后运行到第1个action的prepareAdd时死机报timeout:rollback

我把2个provider各折成一个spring工程，用的是连接nacos，provider的xml如下：

然后我的reference在我的controller里：

我是这样调用的：
@Autowired
private MinusMoneyAction minusMoneyAction;

@Autowired
private PlusMoneyAction plusMoneyAction;

@Override
@GlobalTransactional
public boolean transfer(long from, long to, double amount) throws DemoRpcRunTimeException {

boolean answer = minusMoneyAction.prepareMinus(null, from, amount);
if (!answer) {
// 扣钱参与者，一阶段失败; 回滚本地事务和分布式事务
throw new DemoRpcRunTimeException("账号:[" + from + "] 预扣款失败");
}
// 加钱参与者，一阶段执行
answer = plusMoneyAction.prepareAdd(null, to, amount);

当运行到：boolean answer = minusMoneyAction.prepareMinus(null, from, amount);
eclipse控制台显示：
INFO io.seata.core.rpc.netty.NettyClientChannelManager - will connect to 192.168.56.101:8091
然后我看seata控制台里一直显示：这个globaltransaction超时, will be rollback

我在seata-0.9.0和seata-0.8.1, seata-0.8.0下都是这样。

从昨天下午16：00搞到现在，一分钟没睡，求帮助。

Start with the provider and reference XML shown in the issue, then trace the controller call from prepareMinus to prepareAdd. Check the Seata client connection to 192.168.56.101:8091 and the global transaction timeout or rollback logs. Done means identifying the cause of the timeout and documenting a reproducible configuration or code fix.

Written by the indexing model from the issue text.