IPMI and rescue
Reaching the server when the operating system no longer answers, and what to check before asking for hands in the datacenter.
Out-of-band access is a separate controller on the motherboard, on its own network interface, with its own power. It answers when the operating system does not: a kernel that panics, a firewall rule that locked you out, a boot that stops at the initramfs.
Every dedicated server includes IPMI/KVM access. Request the credentials from the service page in the client area, or by ticket if they are not shown there.
What it gives you
| Action | What it is for |
|---|---|
| Remote console | See the actual screen: boot messages, kernel panic, the login prompt |
| Power cycle | Cut and restore power when the system no longer reboots on its own |
| Virtual media | Boot an ISO you provide, for a custom install or a repair |
| Sensors | Temperatures, fan speeds, power supply state |
Careful
The out-of-band interface is a full path to the machine. Give it a password of its own, never reuse your system password, and do not expose it to a service or a script that does not need it.
Server unreachable: the order to check
- Is it the network or the machine? Run a traceroute from your side, then check the return path from ours on lg.bullionet.com. A route that stops before our edge is not a server problem.
- Does the machine answer at all? Open the out-of-band console. A login prompt means the system is up and the problem is a service, a firewall rule or the routing on the server itself.
- Read the console. A kernel panic, a full disk or a failed mount are all visible there and nowhere else.
- Power cycle only after looking. A reboot that erases the evidence turns a ten-minute diagnosis into a recurring incident.
Rescue environment
When the installed system cannot be repaired from its own console, the server can be booted into a rescue environment: a live system with your disks unmounted, so you can mount them, fix a configuration file, repair a bootloader or copy data out.
Ask for it through a ticket, giving the service identifier and what you intend to do. Our team boots the machine into rescue and tells you how to connect.
Once in rescue, mount your root filesystem before touching anything:
lsblk # identify the root partition
mount /dev/sda2 /mnt
mount --bind /dev /mnt/dev
mount --bind /proc /mnt/proc
mount --bind /sys /mnt/sys
chroot /mnt /bin/bashFrom that chroot, package tooling, grub-install and /etc edits behave as on the running system.
When it really is the hardware
A machine that no longer posts, a failed drive in an array, a power supply that stopped: those need hands in the datacenter. Open a ticket with the service identifier, what the console shows, and when it started. Our engineers are on site in the facilities we operate.
Reviewed on September 1, 2026