quorum write/read bug about firestorm HOT 5 CLOSED

Each partition has replica number of "instances" located on replica number of servers, respectively.
Because the writer works in an asynchronous manner, at each time, "a group of instances" that assigned to a server are issued together.
Therefore, the report should be to send to at least replicaWrite number of servers, which means each server has received this group of instances.

from firestorm.

Each partition has replica number of "instances" located on replica number of servers, respectively. Because the writer works in an asynchronous manner, at each time, "a group of instances" that assigned to a server are issued together. Therefore, the report should be to send to at least replicaWrite number of servers, which means each server has received this group of instances.

举个例子吧，比如说某个 shuffle 操作分配了 5个 shuffle server，然后 spark.rss.data.replica 和 spark.rss.data.replica.write 设置的都是2，那就是说 对于每一个 partition，必须2个副本都report 成功了才算成功是吧，
那也就是说 5个 shuffle server 都必须 report 成功 才行。
但是按照上面的判断逻辑，只要超过2个 shuffle server report 成功的话，就不会抛异常。
不知道我理解的对不对，还请指正，多谢啦！

from firestorm.

Thanks for your issue, there is a PR to fix this.
You could review #107

from firestorm.

Thanks for your issue, there is a PR to fix this. You could review #107

OK, Thanks

from firestorm.

Resovled by #107

from firestorm.