How to Keep a Minecraft Server Online 24/7

Why your server dies the moment you stop looking

A Minecraft server is one program running on one machine. Nothing watches over it, nothing restarts it, and nothing keeps the machine awake on its behalf. When people say "my server keeps going offline", it is almost always one of these five, and none of them are Minecraft's fault:

• The PC went to sleep. The default power settings on every desktop and laptop suspend the machine after a period of no keyboard or mouse activity. A server that is busy serving players still counts as idle, because nobody is touching the keyboard.
• Fast Startup made a shutdown look like a restart. On Windows this is on by default and is the single most confusing one: the machine appears to come back up, but your server process is gone and the startup task you set never ran.
• An update rebooted it at 3 a.m. Windows Update and unattended upgrades on Linux both reboot without asking. If nothing starts your server on boot, it simply never comes back.
• Someone closed the lid, logged out, or unplugged it. Logging out of a user session ends the processes in it. A laptop lid closing suspends the machine. A housemate needing the socket ends the discussion entirely.
• The power or the internet blipped. A one-second cut is enough. Everyone is disconnected, and nothing brings the server back on its own.

What 99.9% uptime actually means

"Always on" is a feeling. Uptime is a measurement, and the distinction matters the moment you try to work out whether your server is actually reliable or you have just not been watching when it dropped.

Real uptime is measured from outside the machine: something on the internet connects to your server's port every minute and records whether it answered. That is the only definition that counts, because it captures the cases your own PC cannot report on — the machine asleep, the internet down, the process hung. "My PC was on all week" is not an uptime figure.

Percentages sound abstract until you convert them into time. Over a 30-day month:

Read that table next to the previous section and the problem with home hosting becomes obvious: a single Windows Update reboot cycle already spends most of a 99.9% month. Not a crisis, not a hardware failure — one routine reboot. Two of those, plus one evening where the machine slept, and you are down at 99% territory without anything having actually gone wrong.

If you want to know your real number rather than guess it, point a free external uptime monitor at your server's address and port and let it run for a month. It will tell you things you did not know, usually about the hours you were asleep. For reference, HytHost runs to a target of 99.9% uptime — the same yardstick, measured the same way.

What a PC running non-stop really asks of you

A machine that never turns off is a different machine from one you use in the evenings. Over a year it accumulates roughly 8,760 hours of runtime — fans spinning, drives turning, heat coming off it continuously. It lives somewhere, and if that somewhere is your bedroom you will learn exactly how audible a case fan is at 3 a.m. in July. None of this is dramatic, but it is the part people forget when they picture "just leaving it on".

It also draws power the entire time, which does show up on a bill. That is a money question rather than an uptime question, so it lives in its own guide — see what a Minecraft server actually costs for the arithmetic.

The bigger surprise is usually not the machine. It is the line it sits behind.

Home connections are built to download, not to serve

Your download speed is irrelevant here. A server sends — a continuous stream of world updates to every connected player — so your upload speed is the ceiling on how many people can play smoothly. Home plans are usually sold with a large download number and a much smaller upload one, and that smaller number is the only one your server cares about. When a server "lags for everyone but the host", this is very often why.

Your address changes without warning

Most home connections get a dynamic IP: it changes when the router reboots, when the line drops, or on the provider's own schedule. Every friend who saved your server in their list now has a dead entry, and you get to send the new address around again. A dynamic DNS service papers over this by giving you a stable hostname that follows the changing address — set one up before you hand the address out, not after.

Some connections cannot host at all

This one catches people after hours of troubleshooting. Many providers place home customers behind carrier-grade NAT, which means you do not have your own public address — you share one. Port forwarding cannot work in that situation no matter how correctly you configure the router, because there is no address to forward from. Others block common server ports outright, or forbid running servers in the terms of the contract. If port forwarding is set up perfectly and outside players still cannot connect, this is the thing to check next.

Your home IP becomes the server address

To let anyone outside your house connect, you forward a port on your router — and from that moment your server's address is your household's public IP address. Everyone you invite has it. So does everyone they pass it to, and anyone who sees it in a screenshot, a Discord message or a server list.

The important shift is what the target becomes. An attack aimed at a hosted server hits that server. An attack aimed at your home address hits the house: your connection, your family's phones, someone's work call, the TV. You cannot filter it from inside, because the traffic has already saturated the line by the time it reaches your router. Turning the server off does not help either — the address is still the address.

This is the one difference between home and datacenter hosting that has no workaround at home. A hosted server's address is disposable and belongs to the server, not to where you live, and it sits behind Voxility DDoS protection that absorbs attacks before they reach it. For a private server shared with people you know, home hosting is a reasonable risk. For anything public, it is the reason most people stop.

Own PC, VPS or managed hosting: who does the work

All three of these can be online around the clock. That is not what separates them. What separates them is who is responsible when they are not — and that is the question worth deciding on, because it is the one you live with every week.

The middle column is the one people misjudge. A VPS fixes the two hardest problems immediately — it has a real address and hardware that never sleeps — but it hands you a bare Linux machine and makes you the system administrator. You install and update Java, you configure the firewall, you write the service that starts the server on boot, you apply security patches, you set up and test the backups. None of that is difficult, but all of it is now yours, permanently, and an unpatched internet-facing machine is a liability rather than a server.

How to keep a self-hosted server alive anyway

If you are staying on your own machine, these are the things that move a home server from "up when I remember" to genuinely reliable. Do them in this order — the first two fix most outages on their own.

Stop the machine going to sleep

Set the power plan so the machine never sleeps and never turns the disks off, set the lid-close action to do nothing on a laptop, and disable hibernation. On Windows, also turn off Fast Startup — it is the reason a machine can look like it restarted cleanly while your startup task never ran. These settings are the difference between an outage every night and none.

Start the server automatically on boot

A reboot you did not plan for should cost you a minute, not an evening. On Linux, a systemd unit both starts the server at boot and restarts it if it exits:

[Unit]
Description=Minecraft server
After=network.target

[Service]
User=minecraft
WorkingDirectory=/opt/minecraft
ExecStart=/usr/bin/java -Xms2G -Xmx4G -jar server.jar nogui
Restart=always
RestartSec=15

[Install]
WantedBy=multi-user.target

On Windows the equivalent is a Task Scheduler task set to run at startup rather than at logon — at logon means nothing happens until somebody signs in, which defeats the point after an overnight reboot.

Add a watchdog, because "running" is not "working"

This is the step almost everyone skips, and it is the one that catches the worst outages. A Minecraft server that has run out of memory or deadlocked is still a running process: systemd sees it alive and does nothing, while players get a timeout. The only reliable test is whether the port actually answers.

#!/bin/bash
while true; do
  if ! timeout 5 bash -c "echo > /dev/tcp/127.0.0.1/25565"; then
    systemctl restart minecraft
  fi
  sleep 60
done

Crude, and effective: every minute it checks whether the server answers on its port, and restarts it if not. Run it as its own service so it survives a reboot too.

Restart it on purpose, before it needs it

Long-running Java servers gradually slow down as memory fragments and plugins accumulate state — the classic "it was fine this morning, it is chugging by evening". A scheduled daily restart at a quiet hour resets all of that in under a minute. Warn players in chat with a countdown so nobody loses a build, and pick an hour when nobody is on.

Back up somewhere other than the machine

A backup on the same disk is not a backup — it is a second copy waiting for the same drive to fail. Copy the world folder on a schedule to a different disk, another machine, or cloud storage. Then do the part nobody does: restore one, once, and confirm the world actually loads. An untested backup is a guess.

Give the address something stable to point at

Set up dynamic DNS so your players connect to a hostname that follows your changing address instead of to a number that expires. It costs nothing and it saves the recurring "the server is down" message that really means "the IP moved".

When the hassle stops being worth it

Everything above genuinely works. Plenty of people run a solid home server for years using exactly those steps. But there are three moments where the effort stops paying for itself, and they are worth recognising when they arrive:

• Someone plays when you are not there. The moment the server needs to be up while you are asleep, at work or away for a weekend, you are on call for it — and a machine in your house cannot be fixed from someone else's kitchen.
• Someone else now depends on it. A world with a few friends' months of building in it is no longer a hobby project. The failure that used to be an inconvenience is now everybody's loss.
• You have restarted it by hand more than twice this month. That is the signal that you are the watchdog, and it does not improve on its own.

The counter-case is just as real, and worth saying plainly: if you play on a LAN, if the machine is on anyway for other reasons, or if running the thing is itself the point, self-hosting remains the right answer and none of the above applies to you.

If you have crossed one of those lines, moving to a managed Minecraft server hands the whole list — sleep settings, boot scripts, watchdogs, scheduled restarts, off-machine backups and DDoS filtering — to someone whose job it is. Plans start small and upgrade in place as the server grows, so you are not deciding your final size today. And if you would rather try a free option first, that is a fair question with its own answer: see the honest look at free Minecraft hosting.