Read the boot console before you reinstall

When a site is down and SSH will not connect, open the boot console before you reinstall. On a VPS that means the panel VNC console; on dedicated hardware it means IPMI or iKVM. The console often shows an fsck prompt, a full disk, or a kernel panic with a useful filename. Those problems are usually short fixes. A reinstall can cost you a day and erase the evidence you needed.

What the boot console shows you

The boot console is the screen you would see with a keyboard attached to the machine. It displays GRUB, filesystem checks, an emergency shell, Windows recovery, or an out-of-memory event that happened before networking started. SSH cannot show any of that, because SSH starts only after those stages succeed. A serial console, when the template offers one, does the same job in text. On dedicated servers, IPMI is the remote console; keep it off the public internet and do not route the website through it. On a VPS, the control-panel console is the equivalent. Test it on a quiet day so you are not learning the interface during an outage.

When a reinstall is optional

A full disk that blocks boot can often be fixed in rescue mode by clearing logs, running fsck, and rebooting. A bad sysctl that hangs networking can be reverted from the console. A Windows update stuck on “getting ready” may only need patience or a last-known-good boot. None of those cases require wiping the guest. Reinstall when the filesystem is truly gone, when the root password is lost and the console cannot help, or when you already wanted a clean template. It should not be the first button you press.

Look at the console, then decide

Open the console and read the last thirty lines or so. If you see fsck, answer the prompt. If you see a panic, take a screenshot. If you see a login prompt, the machine is up and the problem is the network or SSH, not a reason to reinstall. “Server down” is not the same as a stack trace, so treat the console text as the real signal.

df -h
journalctl -xb | tail -n 40
ip -br a

Run those commands on the console when SSH is dead. df -h shows whether the disk is full. journalctl -xb tells the story of this boot. ip -br a shows whether the NIC came up at all. Paste the output into your ticket along with the guest id, a console screenshot, and whether a snapshot exists. If you already reinstalled, say so, because the useful log is gone. A black console is itself the ticket, not a reason for a second rebuild. Put the console URL in your runbook so you are not hunting panel paths at 2 a.m. An offsite copy makes a real wipe survivable; reading the boot console first is often what makes the wipe unnecessary.