Disk health (smartctl)

Dedicated server disks eventually fail. SMART (Self-Monitoring, Analysis and Reporting Technology) reports reallocated sectors, read errors, and SSD wear before total failure. smartctl is part of the smartmontools package.

Prerequisites

• Linux dedicated server with root access
• SATA/SAS/NVMe disks recognized by controller (hardware RAID may hide SMART — see below)
• No destructive tests during production hours without a maintenance window

Installation

sudo apt update
sudo apt install -y smartmontools
sudo systemctl enable --now smartd

Identify disks

lsblk -d -o NAME,SIZE,MODEL,ROTA
sudo smartctl --scan

Example paths:

| Type | Device |

|------|--------|

| SATA | /dev/sda |

| NVMe | /dev/nvme0 |

First SMART read

sudo smartctl -a /dev/sda
sudo smartctl -a /dev/nvme0

Critical attributes (HDD):

| Attribute | Meaning |

|----------|---------------|

| Reallocated_Sector_Ct | Remapped bad sectors — > 0 to watch |

| Current_Pending_Sector | Unstable sectors pending |

| UDMA_CRC_Error_Count | Often faulty cable/SAS |

| Temperature_Celsius | Excessive heat |

SSD NVMe: check Percentage Used, Media Errors, Available Spare.

Short test (non-destructive)

sudo smartctl -t short /dev/sda
# wait ~2 min
sudo smartctl -a /dev/sda | tail -20

Long test (HDD, several hours):

sudo smartctl -t long /dev/sda

Plan a maintenance window — high I/O.

smartd: automatic alerts

Edit /etc/smartd.conf:

/dev/sda -a -o on -S on -s (S/../.././02|L/../../6/03) -m root@localhost
/dev/nvme0 -a -o on -S on -m root@localhost

• -s: scheduled tests (short daily, long weekly)
• -m: email (configure postfix or HolyCloud relay)

sudo systemctl restart smartd
sudo smartctl -i /dev/sda | grep -i smart

Hardware RAID (MegaRAID, etc.)

Physical disk may be /dev/bus/0 — use controller tools:

# MegaRAID example
sudo apt install -y megacli
# or storcli from vendor

Ask HolyCloud support for RAID model for exact smartctl -d megaraid,N -a /dev/sda command.

Quick interpretation

sudo smartctl -H /dev/sda

| Result | Action |

|----------|--------|

| PASSED | Continue monitoring |

| FAILED | Immediate backup, disk replacement ticket |

| Inconsistent data | Cable, backplane, controller |

Logging

sudo smartctl -a /dev/sda > /root/smart-sda-$(date +%F).txt

Keep monthly history to see counter drift.

Dedicated server best practices

• Monitor all RAID disks, not only the visible logical volume.
• Pair SMART with off-server backups (S3, another DC).
• After HolyCloud disk replacement, rerun smartctl -t short.

Troubleshooting

| Problem | Approach |

|----------|-------|

| SMART Disabled | smartctl -s on /dev/sda |

| Device open failed | Disk in RAID without -d option |

| NVMe « unknown» | Update smartmontools |

Need help?

Open a ticket with full smartctl -a output, disk serial (panel / IPMI), and slot for warranty replacement.