Fsck +mdadm don't stop the server to check?

Question

linux

On average, once every 400 days, the server takes offense at disks, IO 100% when deleting a file.
And so on until the demons stop.
and this is basically normal when hundreds of gigabytes and millions of files go through a day. Naturally, the probability of a logical error increases. (I do not take now the problem when the disk really breaks)
There are about 500 sites on the server now.
Raid mirror 4 disks. soft.
The problem is the following, for good, you need to run fsck periodically to check the logical structure of the disks, but of course they have to be unmounted for this.
How to properly deal with this problem? Do not stop the server for professional work for hours like this for 7-10.

Answer 1

I see clearly, the problem is more in a soft mirror raid on 4 disks.
If a mess appears in the FS for no reason, you need to look for a problem (failing memory or disk).

Answer 2

It is logical to try to install ext4 or xfs