Configure Software RAID

What RAID actually protects against

RAID keeps a server running (or lets it be quickly restored) when one physical disk in an array fails. That is the entire scope, it does nothing for accidental deletion, a bad software update, ransomware, or a mistake in a database migration, all of which happily replicate across every disk in the array just as fast as legitimate data does.

RAID 1 vs RAID 10

• RAID 1 (mirroring): two disks, identical copies of everything, half your total capacity, tolerates one disk failing.
• RAID 10 (mirrored stripes): four or more disks, combines mirroring with striping for both redundancy and better performance than RAID 1, half your total capacity, tolerates at least one disk failing per mirrored pair.

For a two-disk dedicated server, RAID 1 is the practical choice. For four or more disks where performance matters as much as redundancy, RAID 10 is generally the better default over RAID 1 with more disks or RAID 5/6, at the cost of the same 50% capacity overhead.

Setting it up

Software RAID on Linux is configured with mdadm, either during OS installation, if the installer offers a RAID partitioning option, or after the fact if the disks are still unpartitioned. If you need RAID configured as part of delivery, request it when you order, since setting it up after data already exists on the disks is a more involved, higher-risk operation than doing it upfront.

Monitoring it

A degraded array (one disk already failed) still works, silently, until the second disk fails too. Set up an alert, "cat /proc/mdstat" or a monitoring agent checking mdadm status, so a single disk failure gets fixed in days, not discovered only when the second disk fails and takes the array down with it.

RAID buys you uptime when a disk dies. It buys you nothing when a person, or a script, deletes the wrong thing. Keep backups regardless of which RAID level you run.