How to configure RAID via BMC on a dedicated server before OS install
Technical guide to configuring RAID 1, 5, or 10 on a dedicated server using the BMC interface (IPMI/iDRAC/iLO) before deploying the operating system.
Configuring RAID via BMC is a fundamental step before installing the operating system on a dedicated server. The RAID controller (PERC, MegaRAID, SmartArray) must have the array created so that the Linux or Windows installer sees a single volume instead of individual disks — otherwise, you install the OS on a standalone disk and lose the redundancy that justified buying the hardware in the first place.
This tutorial is for sysadmins and DevOps engineers preparing a brand-new dedicated server. You will learn how to access the BMC (Baseboard Management Controller — the server’s out-of-band interface), open the remote KVM, enter the RAID controller utility, and create RAID 1, 5, or 10 arrays with sane production settings (write policy, read-ahead, hot spare).
Execution time ranges from 30 to 90 minutes depending on the RAID level chosen and the number of disks — RAID 1 with 2 SSDs initializes in minutes, while RAID 5 with 8 rotational disks can take hours for background initialization to complete. The server remains usable during background initialization, but with reduced performance.
Prerequisites
You need the BMC IP, admin username and password (delivered in the Hostini panel when the dedicated server is provisioned), a Java Runtime or modern browser to access the remote KVM, and the documentation for your RAID controller on hand to check the limits of each level.
https://BMC-IP 443 5900-5910 root or ADMIN Before creating any array, choose the RAID level based on the workload. RAID 1 (mirror) uses 2 disks and loses 50% of the capacity — good for OS and small databases where concurrent reads matter. RAID 5 distributes parity and tolerates a single disk failure — fine for file storage and backups, but avoid it with large SSDs (rebuild can take days). RAID 10 (mirror of stripes) needs 4+ disks and offers the best trade-off between performance and safety for write-heavy workloads (databases, VMs).
Creating a RAID destroys the data on the disks. Confirm that the server is brand new or that you have a full backup before proceeding. The controller does not ask for explicit confirmation in some versions.
Accessing the BMC interface
The BMC web interface runs independently of the main operating system — it is available even with the server powered off, as long as the power cable is connected. That is how you reach the remote KVM and, on some models, configure RAID directly without rebooting.
Open your browser at the BMC IP listed in the Hostini panel:
https://203.0.113.42Accept the self-signed certificate warning — the BMC ships with an internal certificate that is not signed by a public CA. For critical production environments, replace it with a valid certificate later (via the BMC security menu).
Log in with the default credentials shown in the panel. On Dell iDRAC, the user is usually root with password calvin; on HPE iLO, a physical label ships with Administrator + a random password; on Supermicro IPMI, it is typically ADMIN / ADMIN.
Change the default password immediately after the first login — under “iDRAC Settings → Users” (Dell), “Administration → User Administration” (HPE), or the equivalent. BMCs with default credentials are common scanning targets.
Locate the server status on the main dashboard. Confirm that every physical disk is detected — usually under “Storage → Physical Disks” or “System Inventory → Storage”. Missing disks here mean a disconnected cable, a bad slot, or a failed drive.
Opening the remote KVM
The remote KVM (Keyboard, Video, Mouse) gives you access to the server’s screen as if you were physically in front of it. That is how you enter the RAID controller utility during boot.
In the BMC main menu, look for “Virtual Console”, “Remote Console”, or “iKVM/HTML5”. Modern versions use native HTML5 in the browser; older ones require Java Web Start (.jnlp).
iDRAC: Console → Launch Virtual Console
iLO: Remote Console → HTML5 Console
IPMI: Remote Control → Console Redirection → LaunchReboot the server through the BMC (not through the OS):
Power → Power Cycle System (cold boot)A power cycle forces a clean boot needed to reach the RAID controller during POST. A normal reboot may not reach that screen on some BIOSes.
Entering the RAID controller utility
During POST (Power-On Self Test), the RAID controller shows its banner for a few seconds and accepts a key combination to enter configuration. You need to be alert — the window is short.
Watch the KVM screen right after the reboot. When the controller banner appears, press the matching combination:
| Controller | Combination | Typical banner |
|---|---|---|
| Dell PERC (LSI/Broadcom) | Ctrl+R | ”PERC H750 BIOS” |
| HPE Smart Array | F8 | ”HPE Smart Array Px-y” |
| LSI/Broadcom MegaRAID | Ctrl+H or Ctrl+R | ”MegaRAID BIOS” |
| Lenovo ThinkSystem | F1 → Storage | UEFI menu |
Modern servers in UEFI mode show controller configuration inside the System Setup menu (F2 or DEL at boot), instead of the legacy Ctrl+R screen.
Inside the utility, identify the detected physical disks. They appear as “PD 0”, “PD 1”, etc., with size, model, and state (Ready/Unconfigured Good). Disks marked as “Foreign” carry an old RAID configuration from another server — clear them with “Clear Foreign Config” before use.
Creating the RAID array
With disks identified and cleaned, the next step is to create the virtual disk (VD) — the logical volume the OS will see.
Start VD creation:
Create Virtual Disk → RAID Level: [choose 1, 5, 10]Select the physical disks that will be part of the array. For RAID 1, two disks. For RAID 5, a minimum of 3 disks (4-6 recommended). For RAID 10, a minimum of 4 disks, in even numbers.
Configure the VD parameters with sane production values:
64 KB or 256 KB Read Ahead Write Back (with BBU) Disabled Fast Initialize A 64 KB strip size is good for mixed workloads; 256 KB favors large sequential IO (video, backup). Read Ahead speeds up sequential reads at little cost. Use Write Back only with a healthy battery/capacitor — otherwise, keep Write Through.
Enabling Write Back on a controller without an operational battery or capacitor exposes data to loss during power failure. The controller acknowledges writes to the OS before committing them to disk — if the server powers off at that instant, the data sits only in volatile DRAM.
Confirm the creation. The utility warns that the disks will be initialized. Accept and wait for the fast initialization (seconds) to complete. Full initialization (background init) continues after the OS boots — you can use the array normally, but with degraded IO until it finishes.
If the server has an extra disk (for example, 5 disks for RAID 5 + 1 hot spare), assign the leftover disk as Global Hot Spare. If any disk in the array fails, the controller starts the rebuild automatically with no human intervention.
Verifying the array
Before installing the OS, confirm that the array was created correctly and is visible in BIOS/UEFI.
Exit the controller utility (Esc → Save Configuration → Yes) and let the server continue booting. Enter the boot menu (F11 or F12) and confirm that the virtual disk appears as an available boot device — it will be listed as “PERC H750 VD 0” or similar, instead of the individual physical disks.
Go back to the BMC, under “Storage → Virtual Disks” (or equivalent), and confirm that the VD shows “Optimal” status and that initialization progress is visible. If it shows “Degraded” before any usage, a disk is faulty — investigate the controller logs.
After that, mount the operating system ISO through the BMC Virtual Media and proceed with the normal install. The installer will see the RAID as a single disk and partition over the logical volume.
Troubleshooting
Disks appear as “Foreign”
Disks coming from another server keep the old array signature. In the controller utility, pick “Foreign Config → Clear” to remove it. Do not use “Import” unless you are recovering a complete array from another machine.
Controller battery in “Charging” or “Failed”
Write Back is automatically disabled until the battery charges (up to 4 hours on first use) or is replaced. Use Write Through temporarily — write performance drops considerably, but data stays safe. BBU replacement is a maintenance service; open a ticket with Hostini.
Background init too slow
The rebuild/init competes for IO with the normal workload. In the controller utility, adjust “Rebuild Rate” between 30 and 60 — lower values prioritize the application, higher values speed up the rebuild but degrade performance. For critical production, keep it at 30 and tolerate the extra time.
Next steps
With RAID configured and the OS installed, there are related topics that deepen dedicated server management:
- Configure controller health monitoring via SNMP or vendor agent (OMSA, AMS, MegaRAID Storage Manager) to get disk failure alerts before they become visible degradation.
- Document the RAID scheme, disk serial numbers, and slot positions in your internal inventory — it makes correct replacement easier in case of failure.
- Test the hot swap procedure before you need it: intentionally pull a disk, confirm the array degrades and rebuilds with the spare, reinsert it, and validate again.
- Evaluate at-rest encryption via SED (Self-Encrypting Drives) with key management on the BMC for environments with regulatory requirements.
If you are building a new fleet and want hardware with a modern BMC, a healthy BBU, and disks in hot-swap slots, Hostini’s dedicated servers ship with a manageable controller and a BMC with HTML5 KVM — ready for you to configure RAID before deployment.
Frequently asked questions
What's the difference between hardware RAID via BMC and software RAID (mdadm)?
Hardware RAID is configured on the physical controller (PERC, MegaRAID, SmartArray) and presents a single volume to the operating system, with dedicated cache and battery. Software RAID (mdadm) is managed by the Linux kernel, depends on the CPU for parity, and requires individual disks visible to the OS. Hardware offers better performance on RAID 5/6 and survives motherboard swaps; software is more portable and cheaper.
Can I change the RAID level later without reinstalling the server?
Some controllers support online level migration (OCE — Online Capacity Expansion or RLM — RAID Level Migration), but the process is slow (hours to days) and risky for production. The safe path is to back up, destroy the current array, create the new level, and restore. That's why choosing the level before deployment is critical.
Do I need a battery (BBU) or capacitor (CacheCade) on the controller to use write-back?
Yes. Enabling write-back without a battery or capacitor puts data at risk during power loss — any write still sitting in the controller's DRAM cache is lost. Without cache protection, keep write-through. Check in the controller utility that the battery status is Optimal before switching to write-back.
How do I know which BMC vendor my server uses?
The vendor depends on the server brand: Dell uses iDRAC, HPE uses iLO, Supermicro uses generic IPMI or SuperBMC, Lenovo uses XCC. Hostini lists the type in the dedicated server panel, along with the BMC IP, username, and password. The RAID controller model (PERC H750, MegaRAID 9560, etc.) also appears in the inventory.
Does a global hot spare cover all arrays or only the array it was assigned to?
A global hot spare covers any array on the controller that suffers a disk failure. A dedicated hot spare only serves the specific array it was bound to. On servers with multiple arrays (e.g., RAID 1 for OS + RAID 5 for data), the global option is more efficient — a single spare disk protects every volume.
Is it possible to configure RAID via BMC without remote KVM, using only the command line?
Yes. Utilities such as racadm (Dell), HPE iLO REST API, and StorCLI (LSI/Broadcom) let you create and manage arrays over SSH or HTTPS on the BMC interface, without needing the graphical KVM. The graphical path is more instructive for the first time; the CLI path is faster for automation.