The replication queue is too high due to the NIC of the Exchange server!

The client has reported a fault recently. One of the problems mentioned is that the mailbox backend database mailbox has a high replication queue performance. Although it will be automatically digested, it takes a long time to process it. From the perspective of the queue type, the replication queue has a certain relationship with network replication, rather than the replay log problem. Therefore, the disk I/O problem can be basically ruled out! Therefore, we basically set the problem to the problem caused by the NIC problem. After careful consideration, we can see that not all servers have this situation, but occasionally one or two servers appear.

After my judgment, I could not confirm the problem. I started the problem at Microsoft in case and was diagnosed by Microsoft engineers, it is found that there are several ways that do not comply with exchange BPA:

1. The NIC uses the teaming function, which is not supported in Microsoft official documentation, but is still supported in actual situations.

2. the toe function of the network card is enabled. The TOE function is a good function in use as a memory card. However, some of Microsoft's standard services, such as Ad \ exchange \ Moss \ SQL, we recommend that you disable this part of these server products!

 

After Microsoft engineers started the teleconference, they discussed the result that they removed the NIC teaming function, disabled the NIC toe function, and then performed the teaming function. The problem was solved.