Windows Failover cluster tuning

Microsoft official document: https://docs.microsoft.com/en-us/previous-versions/windows/it-pro/windows-server-2012-R2-and-2012/dn265972(v=ws.11)

This document is quite detailed, and the official instructions of Microsoft are all in English. Just do it. One of the problems I am talking about today is a case study.

Physical structure, two knife boxes, each with three blades and one storage. The four blades of the two knife boxes make up a failover cluster. On this cluster, two mail servers are running, two servers are DAG, and of course there are other applications floating around. The failure phenomenon is to find that the running efficiency of the two mail servers is a little inefficient, and I often time out when I perform cross-forest data migration.

Troubleshooting:

Blade 2rec 3pm 5pm 6, the connection to the storage is enabled with multipathing, in other words, any virtual machine can easily access the two storage volumes. But the current problem is that two mail servers are floating on the above two blade physical nodes, causing the data traffic of one of the mail servers to pass through the network between the two blades.

Transformation method:

1. On the cluster manager, migrate one of the mail servers online to another knife box physical host.

two。 On the mail server, hang different mail databases on two different mail servers, not all in one, this kind of operation is graphical, very simple.

3. To tune the mail server, refer to the link below, modify the timeout of the two mail servers across the forest, and restart the mailbox replication service.

Https://social.technet.microsoft.com/Forums/en-US/279bac54-f7bd-4d5e-884e-6c8701db13cc/remote-2010-2010-crossforest-mailbox-move-timeout?forum=exchangesvrdeploylegacy

The main idea to solve the problem is to balance the load and try not to cause cross-network access. Although there should be no problem in theory, it will form a bottleneck in practice. To deal with the timeout problem, because it can not change the remote network delay, we can only change the timeout parameter. It's a way anyway, it's better than doing nothing.