Getting started

Dedicated server not responding: what to check and report

If your website, SSH or RDP stopped opening, in 15 minutes you can tell whether the server is down or the network is in the way, reboot it yourself and collect everything support will ask for.

  • Time: 15 minutes
  • Level: Intermediate

In short

  • Start on your side: ping the server and check the service’s port, then do the same from another network, such as mobile internet. If the server responds from the other network, the cause is in the network between you and the server or in the firewall on the server.
  • Check your payments and your e-mail. The server is suspended as soon as the paid period ends if your balance was short, and access to it is restricted if a complaint goes unanswered for 24 hours.
  • Reboot the server yourself: on the Reset tab in Hetzner Robot or with the Restart button in the OVHcloud Control Panel. If that does not help, look at the screen through the console and boot into Rescue.
  • In your message to support, give the server, since when and what is not working, what you have already tried and whether we may reboot it, and attach the mtr output to the server.
Contents
  1. Check whether the server responds and the port is open
  2. Repeat the check from another network
  3. Run an mtr trace: support will ask for it
  4. Check whether the server was suspended for non-payment or a complaint
  5. Reboot the server from the provider’s panel
  6. Look at the server’s screen through the console
  7. Boot into Rescue if the system does not start
  8. How to tell that the server is working again
  9. The server is still down: what to send support
  10. FAQ

You will need

  • The server’s IP address: it is in the server delivery ticket.
  • A Hetzner Robot login from ABCD or your own OVHcloud account, to reboot the server.
  • A command line on your computer: PowerShell on Windows, Terminal on macOS and Linux.

Start with checks on your own computer: they take a couple of minutes and often show the cause right away. Rebooting, the console and Rescue are for when the server does not respond from anywhere. The examples use the server address 203.0.113.10; replace it with yours.

Check whether the server responds and the port is open

You need two checks. ping shows whether packets reach the server, and a port check shows whether the service itself accepts connections. Port 22 is SSH, 3389 is RDP, 443 is a website over HTTPS.

Windows. Open PowerShell: press Win, type powershell and press Enter. The first command sends the server four packets and shows how many came back:

ping 203.0.113.10

The second tries to connect to the port. Look for the TcpTestSucceeded line at the end of the output: True means the port is open, False means it is not.

Test-NetConnection 203.0.113.10 -Port 22

macOS and Linux. Open Terminal. The first command sends four ping packets:

ping -c 4 203.0.113.10

The second tries to connect to port 22 and waits 5 seconds for an answer. On macOS the timeout is set with the -G option:

nc -vz -G 5 203.0.113.10 22

On Linux it is the -w option. If there is no nc command, install the netcat-openbsd package on Ubuntu and Debian or nmap-ncat on AlmaLinux and Rocky Linux.

nc -vz -w 5 203.0.113.10 22

The word succeeded or Connected in the reply means the port is open. Connection refused means the server is up, but nothing is running on that port or a firewall rejects the connection. A timeout (timed out, TIMEOUT) means the connection did not reach the server.

PingPortWhat it meansWhat to do
RespondsOpenThe server and the network work; the failure is in an application: a website, a database, a game serverLog in to the server and check that application’s logs
RespondsClosedThe server is on, but the service is stopped or the port is closed by a firewall: on the server itself, on the Firewall tab in Robot or in the OVHcloud Edge Network FirewallLog in through the console, start the service, check the firewall rules
No responseOpenThe server works, and ping is blocked by a firewall. Windows Server is often set up this way by defaultNothing to fix; check by the port
No responseClosedThe server is off, has frozen, or the network does not reach itGo on to the next steps

Sometimes the server responds on its IP, but the website does not open by name. Then check the name: nslookup server.example.com works on all three systems and should return your server’s IP.

Repeat the check from another network

Connect your computer to the internet through your phone, or ask someone in another city to check ping and the port. The result tells you where to look next.

  • It responds from another network, but not from yours. The server is working, and the connection is lost along the way. The cause can be at your internet provider or on the server itself: its firewall may have blocked your IP, for example fail2ban after several wrong passwords. Run a trace as described below.
  • It fails only from networks in one country. In some countries, local authorities restrict access to some IP addresses (clause 4.4 of the offer).
  • It does not respond from anywhere. You can skip the trace: go on to checking payment and rebooting.

Write down which networks you checked from and what you got: support will need this.

Run an mtr trace: support will ask for it

A trace shows every hop on the way to the server and the packet loss at each one. It shows where the connection breaks: at your provider, in a backbone network or at the data center. Run it from the computer the server does not open from.

Windows. The built-in pathping command does the same as mtr: it first finds the hops, then counts the loss at each one. It runs for several minutes; wait for the final table.

pathping 203.0.113.10

If you need a quick result, tracert 203.0.113.10 shows the hops along the way within a minute, but without the loss.

macOS. mtr is installed with Homebrew and run with sudo: the system will ask for your Mac password. Without Homebrew, the built-in traceroute 203.0.113.10 will do, but it does not count loss.

brew install mtr
sudo mtr -rwc 100 203.0.113.10

Linux. Install mtr: the first command is for Ubuntu and Debian, the second for AlmaLinux, Rocky Linux and Fedora. The third runs the trace.

sudo apt install mtr-tiny
sudo dnf install mtr
sudo mtr -rwc 100 203.0.113.10

The -r option prints a report instead of a live screen, -w keeps host names from being cut off, and -c 100 sends 100 packets to each hop. The report appears after about two minutes and looks like this:

HOST: laptop                 Loss%   Snt   Last   Avg  Best  Wrst StDev
  1.|-- 192.168.1.1             0.0%   100    1.1   1.3   0.9   4.8   0.5
  2.|-- 198.51.100.1            0.0%   100    5.0   5.4   4.2  11.9   1.0
  3.|-- 198.51.100.17          38.0%   100   21.9  22.6  21.5  29.8   1.3
  4.|-- 192.0.2.33              0.0%   100   38.4  38.8  37.9  44.7   1.1
  5.|-- 203.0.113.10            0.0%   100   39.0  39.3  38.5  43.6   0.8

First look at the Loss% column in the last line: that is the server itself. In the example it shows 0%, so the connection is fine, and the 38% at the third hop means nothing: routers often answer these packets last. The problem is where the loss starts and continues to the end of the list. If the last line shows 100% or there are ??? instead of addresses, packets do not reach the server. Copy the report from the terminal window and send it to support as text.

Check whether the server was suspended for non-payment or a complaint

If your balance did not have enough funds to renew at the end of the paid period, the server is suspended immediately (clause 6.4 of the offer). The end date is shown in the my.abcd.host client area, and the exact date of data deletion is in the payment reminder.

Data is deleted permanently: for OVHcloud servers 5 days after suspension, for Hetzner and other data centers after 24 hours. If you pay before that, the server is restored within 24 hours of payment. Details are in the article on renewal and suspension.

The second reason is a complaint (abuse report). Complaints about the server are sent to the e-mail address in your client area and to a ticket. If you do not reply and remove the violation within 24 hours, access to the server or IP address is restricted; the data center may set a shorter deadline, and it is stated in the message (clause 4.5.1 of the offer). If attacks, scanning or spam come from the server, access may be restricted immediately, without notice, and we tell you the reason within 2 hours after blocking (clause 4.5.2). What to do about a complaint is covered in the article on abuse complaints from Hetzner and OVH.

It can also be the other way round: your server is under attack. During a DDoS attack we may limit bandwidth or temporarily null-route the IP address (clause 5.2).

Reboot the server from the provider’s panel

If you can still log in to the server, reboot it from the system: sudo reboot on Linux or Start → Power → Restart on Windows. If you cannot, reboot it from the panel.

A hardware reset from the panel is the same as pulling the power. Programs do not get to save their data, and a disk check may run at boot. If the server responds over SSH or RDP, reboot it from the system.

  • Hetzner: in Hetzner Robot, open Serveryour serverReset. Which reset type to choose is covered in the article on rebooting in Robot.
  • OVHcloud: on the server page, on the General information tab, in the Service status block, click “...” next to Status, then Restart and Confirm. Details are in the article on rebooting in the OVHcloud Control Panel.

Robot or the OVHcloud panel will not open? Your internet provider sometimes restricts access to them. Tell us on Telegram or in a ticket which server to reboot, and we will reboot it ourselves.

Booting takes a few minutes, and longer on servers with a lot of memory and many disks. Do not repeat the reset right away: wait 5–10 minutes and check ping and the port again.

Look at the server’s screen through the console

If the server still has not responded after the reboot, you need to see what is on its screen. A remote console shows the screen and gives you a keyboard, even when the network on the server is down.

  • Hetzner: a data center technician connects the KVM console, and we order it when you ask in a ticket. The first 3 hours are free. It is not connected instantly, so ask for the console as soon as it is clear that the reboot did not help: Hetzner KVM console.
  • OVHcloud: open the IPMI/KVM tab on the server page in your panel: Rescue and IPMI on an OVH server. Kimsufi servers have no IPMI.
What is on the screenWhat it meansWhat to do
A login: prompt or the Windows sign-in screenThe system has booted, but the network or a service has not come upLog in at the console and check the network with ip -br addr
Give root password for maintenanceThe system has stopped at a file system check or an error in /etc/fstabEnter the root password and look at the error, or boot into Rescue
No bootable device or a blank screen after the BIOSThe server did not find a disk to boot fromBoot into Rescue and check the disks
Kernel panicThe kernel failed to load, often after an updateChoose the previous kernel in the GRUB menu or repair the system from Rescue

The network is gone after a hardware replacement. After a motherboard or network card is replaced, the network interface may get a different name, and the system does not bring up the network with the old settings. At the console, ip -br link shows the new name. Put it in place of the old one in the network settings (the files in /etc/netplan/ on Ubuntu or /etc/network/interfaces on Debian) and reboot the server. Without a console, you can do the same from Rescue: once the partition is mounted, these files are in /mnt/etc/.

Whatever is on the screen, a screenshot of it in the ticket helps support understand faster what happened.

Boot into Rescue if the system does not start

Rescue is a small system that boots over the network into the server’s memory. It does not touch the disks. From it you can check the disks, read logs, fix settings and copy files.

  • Hetzner: the Rescue tab in Robot, then a reset on the Reset tab. The activation is valid for one boot and expires if you do not reboot the server within 60 minutes: Hetzner Rescue System.
  • OVHcloud: the boot mode is changed next to Boot on the General information tab, and then the server is rebooted with the Restart button. When you are done, switch back to booting from the disk, or the server will boot into Rescue again: Rescue and IPMI on an OVH server.
  1. Log in to Rescue via SSH

    Connect as usual, as the root user: with the password Robot showed you or OVH sent you, or with an SSH key if you chose one when you enabled Rescue. The SSH client will warn you that the server key has changed: Rescue has its own key, so this is expected.

  2. Look at the disks and RAID

    The first command shows the disks and partitions, the second the state of software RAID.

    lsblk
    cat /proc/mdstat
  3. Mount the system partition

    The command mounts the system partition at /mnt. Take the name from the lsblk output: it is usually the largest partition or array. On Hetzner servers with software RAID it is often /dev/md2, on OVH for example /dev/md3, and without RAID a partition such as /dev/sda3 or /dev/nvme0n1p3.

    mount /dev/md2 /mnt
  4. Read the system log

    The command shows the last 50 errors from the journal of your system, not of Rescue.

    journalctl --directory=/mnt/var/log/journal -p err -n 50

    If you get No journal files were found, the log is kept in plain files. tail -n 100 /mnt/var/log/syslog shows the last lines on Ubuntu and Debian, and tail -n 100 /mnt/var/log/messages on AlmaLinux and Rocky Linux.

If the RAID has fallen apart or a disk is throwing errors, go to the article on checking disks and RAID.

How to tell that the server is working again

Ping gets through without loss, the port check shows True or succeeded, and the service responds: the website opens, SSH or RDP lets you in. If the server came back after a reset, look at the logs and find out why it froze: otherwise it may happen again.

The server is still down: what to send support

Write in a ticket or on Telegram. Support will check the server from our side: whether it is reachable, its status at the data center and whether any maintenance is going on there. So that support does not have to ask you again, copy the template and fill it in:

Server: #1234567, IP 203.0.113.10
Not working since: 21 November 2026, 14:30 UTC
What does not work: ping, SSH and the website do not respond
Checked from: home internet and a mobile network
Already tried: hardware reset in Robot at 14:45, did not help
OK to reboot for diagnostics: yes
mtr trace: attached

The reboot line saves time: without it, the support engineer will first ask for your permission. The mtr output is needed if the server does not open only for you: it shows at which hop the packets are lost. If support asks for a trace in the opposite direction, it is run on the server with the same mtr command, only to the external IP address of your computer.

If the problem has to go to the data center’s engineers, the response time depends on them and is usually longer at weekends. We pass their replies on in the same ticket.

FAQ

My Hetzner server was delivered in Rescue, I rebooted it, and now it does not boot. What happened?
If a server is delivered with only the Rescue System, there is no system on its disks yet: Hetzner auction servers, for example, are delivered this way. After a reboot it has nothing to boot from. Install an OS: reinstalling Linux in Robot or installing with installimage.
Will a reboot or Rescue erase my data?
No, the files on the disks stay. A hardware reset can lose only what programs had not yet written to disk, and Rescue loads into the server’s memory and does not change the disks. Data is erased by an OS reinstall.
Support asks for mtr, but I use Windows. What should I send?
The output of the built-in pathping command with your server’s address: it also shows the loss at each hop. Wait for the final table and send the whole output as text.
Robot will not open, and I need to reboot the server. What can I do?
Your internet provider sometimes restricts access to the panel. Tell us on Telegram or in a ticket which server to reboot, and we will do it ourselves.

Did this guide help?

Questions?