Cluster on Server 2012?

O

other_letter2015-12-08 11:17:17

System administration

other_letter, 2015-12-08 11:17:17

And here's how to do clusters right now?
I look and see that two servers are meant, which look at the network with one interface, and with the second - into the library where the virtual machines lie.
Of course, "one interface" is all conditional.
From my understanding, the cluster is made (among other things) to continue working when one of the servers fails. It seems that everything is correct according to the scheme. But is the library really considered so reliable?
In general... Confused. How is it right now?

Reply

Answer the question

In order to leave comments, you need to log in

2 answer(s)

I

Ivan, 2015-12-08
@Amigo83

Shared storage that is configured for RAID is considered trusted. The general scheme of a failover cluster is as follows: two servers are connected to a common storage (fiber channel, iSCSI) OS is installed on the server disks. Both servers are connected to a shared storage, on which a special volume is marked - cluster shared volume, on which VHD disks will lie. And now both servers can use these disks, that is, if one server fails, there is a transparent switching between the nodes, and the virtual machine continues to run on the other.

V

Vladimir Zhurkin, 2015-12-08
@icCE

>I look and see that two servers are meant, which look at the network with one interface, and with the second - into the library where virtual machines lie.
Often this method is used for examples, it's true. But in real work, I do not recommend using two servers. In this case, the storage system can participate in the quorum. But it's almost a splitbrain and I'd recommend just stopping the work rather than keep using it like that.
> From my understanding, the cluster is made (among other things) to continue working when one of the servers fails. It seems that everything is correct according to the scheme. But is the library really considered so reliable?
SHD is certainly reliable, but not so. Typically, a storage system has two heads (controllers) and when creating a cluster, it can take part in a quorum. You can raise RAID on storage systems or use not storage systems at all, but the so-called JBOD baskets. And do all the partitioning and RAID using Windows tools (starting from 2012).
At the same time, if you are going to use virtual machines, I would recommend raising a continuous replication server, which will allow you to rollback a virtual machine very quickly.
By the way, storage systems themselves can be in the mirror and storage systems can be created for continuous replication.
When creating a cluster, you also need to remember about switches that can fail, and then switches come that can stack with each other.