--- title: RAID description: "Combine multiple disks for redundancy or performance. RAID is not backup, and rebuilds carry risk." canonical: https://penasihathosting.com/en/hosting-wiki/raid type: wiki locale: en updated: 2026-10-29 author: Willya Randika --- # RAID ## Overview - **Summary:** Combine multiple disks for redundancy or performance. RAID is not backup, and rebuilds carry risk. - **Author:** Willya Randika ([profile](/penulis/willya-randika)) ## Article RAID combines multiple disks into one volume to improve redundancy, performance, or both. RAID can keep a service running when one disk fails, but it does not protect against deletion, ransomware, or a whole-server failure. RAID is not backup. ## Common levels RAID 0 uses two or more disks for striping. It is fast but has no redundancy. RAID 1 mirrors two disks, so effective capacity is 50 percent. RAID 5 uses parity and survives one disk failure. RAID 6 uses dual parity and survives two disk failures. RAID 10 combines striping and mirroring, usually with better performance and rebuild behavior. RAID 5 on large disks is risky during a rebuild. Reading the entire array can take a long time and trigger a second disk failure. For important data, RAID 6 or RAID 10 is often wiser. ## Rebuilds and hot spares When a disk is replaced, the array rebuilds. During that process, performance can drop and the other disks carry extra load. A hot spare is a standby disk that automatically replaces a failed one, speeding recovery at extra cost. Check status with `cat /proc/mdstat`, `sudo mdadm --detail /dev/md0`, and `sudo smartctl -a /dev/sda`. Watch for clean state, active device count, and SMART media errors. Do not wait for degraded status. ## Hardware and software RAID Software RAID such as mdadm, ZFS, or Storage Spaces uses CPU and is more portable. Hardware RAID uses a controller with cache and battery or flash backup. Performance is good, but configuration and controller replacement can be painful. Cloud providers often provide redundant volumes underneath, so RAID inside a VM may be unnecessary. Check the SLA and volume configuration before adding complexity. ## Common mistakes Treating RAID as backup is the most dangerous mistake. A deleted file is still deleted on every disk. Not monitoring SMART also lets a disk fail without an application warning. Mixing disk sizes or speeds without planning makes capacity follow the smallest disk and performance uneven. A backup that has never been restored is unproven. RAID relates to [NVMe SSD](/en/hosting-wiki/nvme-ssd), [hosting backups](/en/hosting-wiki/hosting-backups), and [block storage](/en/hosting-wiki/block-storage). For a whole-server snapshot, see [VPS Snapshot](/en/hosting-wiki/vps-snapshot). ## FAQ RAID 1 survives one disk failure, but not deletion, malware, or server damage. You still need offsite backups. It depends on disk size, speed, and load. It can take hours or more than a day. Never assume a rebuild finishes unattended. Often not, because the provider already provides redundancy underneath. Check the SLA and volume configuration before adding complexity.