How reliable are Minecraft servers? Uptime, measured

Most Minecraft servers that are actually running stay up almost all the time, but a large share of listed servers are not running at all. Over the last 30 days the median server that was playable at least once was up 98.9% of the time, while 83 of the 212 listings we measured (39%) could not be joined even once.

Data from our own pinger over the last 30 days. Hours are shown in your time zone. Every figure on this page is live and recalculated from our latest pings.

Headline numbers

Right now 100 listings are playable. Over the last 30 days:

39%Listings that were never playable in the last 30 days

Measured 19 Oct 2026, 22:55 UTCHow we measure

The two 99.9% rows are the reason this report exists. Count every answer as "online" and 52 listings look near perfect. Count only answers from a server you could join and the figure drops to 33.

Responding is not the same as playable

When a list shows a green dot, it usually means the address answered a status ping, the same request your multiplayer menu sends to draw the server row. Plenty of things answer that request without being a game you can join. Answers our pinger has received, anonymized:

• Version text "OFFLINE" and a message telling you to join to start the server, from a free host's sleep page
• "Sleeping, connect to wake"
• "Server no longer exists"
• "Currently under maintenance, join Discord for updates"
• "Server not found", sent by a hosting company's lobby for an address that no longer points at anything

Each of these is a perfectly good ping response. None of them puts you in a world. So we count a ping as playable only when the server answers, offers more than zero player slots, and has no placeholder wording in its version text or MOTD. The full rule and the word list are on the methodology page.

The gap is not small. 11.0% of all the answers we received in the last 30 days were placeholders, and some listings answer nearly every ping without being joinable once. A list that counts answers ranks those among its most reliable servers.

The uptime ladder

• Never answered65
• Placeholder only18
• Under 50%25
• 50 to 90%21
• 90 to 99%19
• 99 to 99.9%31
• 99.9% or more33

83 of 212 measured listings were never playable. 33 were playable at 99.9% of our pings or more.

Measured 19 Oct 2026, 22:55 UTCHow we measure

What each rung of the chart means in hours: a 30-day month has 720 hours, so 1% downtime is about 7 hours and 0.1% is about 43 minutes.

So what is good uptime for a Minecraft server? From this data, 99% is a fair bar for a public server, and 64 of the 129 running listings cleared it over the last 30 days. 99.9% is excellent. It allows about 43 minutes of downtime a month, so a daily restart that is back within a minute fits, and a slower one does not. Anything under 90% means a player who tries a few times a week will find it closed at some point.

Outages: how long and when

To look at outages without dead listings drowning the numbers, we take the listings that were playable at least 90% of the time, 83 over the last 30 days, and log every run of three or more missed pings in a row that ended with the server coming back.

• There were 472 such outages in the last 30 days.
• The median outage lasted 5 minutes.
• 70% of them were over within 15 minutes.
• During many outages the address keeps answering, with zero player slots or a placeholder message, instead of going silent. A list that counts answers never sees those outages.

Short outages are common and cheap. The few that run for hours are what players remember: the difference between "it was restarting" and "it was gone all evening". Every server page on serverlist.net keeps its own outage log, so you can tell which kind a server has.

When do they start? Over the last 30 days the most common start hour for short outages was 04:00 UTC. Treat that hour as a hint rather than a rule: one server that restarts at the same time every day adds about 30 outages to its hour on its own. Scheduled restarts also show in the minutes. In the 30 days to 19 October 2026, about one short outage in six began in the first four minutes of an hour, more than twice the share chance would give. Short outages hit peak time too: 19% of them started in the four busiest hours of the network (see when Minecraft servers are busiest), which hold one sixth of the day. If you run a server, our guide to finding the cause of lag before you fix it covers when a restart helps and when it only hides a problem.

Peak 15:00 UTC

0003060912151821

Times in UTC

Busiest at 15:00 UTC with about 18,099 players, quietest at 01:00 UTC. The busiest hour holds 2.6 times the players of the quietest.

Measured 19 Oct 2026, 22:55 UTCHow we measure

By region

Owners declare the countries their server is for, and we group a listing into a region when all its countries fall in one. Last 30 days:

Listings that declare several regions, or none, are left out of this table. Regions with only one or two listings are too small to say anything and are not shown.

The Asian rows need a caveat. Our pinger checks from one location in western Europe. The fastest answers from European servers take a median 10 ms, from North American ones 78 ms and from Asian ones 147 ms. A long route that drops packets can turn a slow answer into a missed ping, so part of a gap may be the path between us and the server rather than the server itself. Distance is not the whole story, though, which is why North American servers can be far away and still have the best record. Treat the Asian rows as a lower bound on how reliable those servers feel to players nearby.

Free hosts and paid hosts

We cannot see who pays for hosting. The address gives a hint, though. Free hosts such as Aternos and Falix give every server a subdomain of their own, stop the server when nobody is on, and answer pings with a sleep page until someone starts it again.

Over the last 30 days, 5 listings used such an address. Our pinger got some answer on 64.5% of its attempts, but a joinable server sat behind that answer on only 6.2%. The other listings that were playable at least once had a median playable uptime of 99.0%. The free host group is small, so read it as an example rather than a rate. It costs nothing, and it also means a passing player almost always finds the server asleep.

Tunnels and dynamic DNS names usually point at a computer at someone's home, so uptime follows that computer. An own domain does not make a server reliable either: most never-playable listings have one, probably servers whose owners moved on. If you are choosing where to host, our guide on choosing Minecraft server hosting goes through what matters beyond the price.

How uptime affects ranking on serverlist.net

Uptime is part of how the list orders servers, and placeholders count as down everywhere:

• Listings that were playable at least once in the last 24 hours sort above all others. Within each group, the order is votes in the last 30 days, then all-time votes, then the newest listing.
• Every row in the list shows 7-day uptime. Every server page has a 90-day uptime calendar, and its stats page keeps an outage log, so you can see the pattern before you join.
• A server page that has not been playable for 14 days says so in a notice, and we ask search engines not to index it until the server is playable again.

The reason is simple. On a list ranked by votes alone, a dead listing that still collects votes from a loyal Discord can sit above servers you could join right now. Ranking playable servers first keeps the top of the list joinable. If you are comparing servers, uptime is one of the signals in how to choose a Minecraft server worth your time.

Is it down, or is it you?

If a server you play keeps going offline, check it from outside first. The server status tool pings it from our side and tells you whether it is online, offline, asleep on a free host or in maintenance.

If it is online for us and still fails for you, the cause is usually on your side of the connection, and can't connect to a Minecraft server maps each error message to its usual causes.

How we measured

• Pings. Every public listing is pinged with the Java status ping on its Java address, and crossplay listings also get a Bedrock ping on their Bedrock port. Each listing is checked every minute, and before 17 October 2026 every 70 to 90 seconds.
• Playable. An answer counts as playable when it offers more than zero player slots and carries no placeholder wording in the version text or MOTD. Everything else that answers is "responding but not playable".
• Uptime. Playable pings divided by all pings for that listing in the window. A listing added during the window counts from its first ping. Hours in which our pinger did not run count as neither up nor down.
• Outages. Three or more consecutive pings that were not playable, ending with a playable ping, among listings playable at least 90% of the window.
• Exclusions. Listings that sit on a shared hosting lobby or point at another network's address are left out, 4 of 216 public listings right now. Every other listing counts on its own, so a network listed twice is counted twice.
• Blind spots. We see the internet from one place. A proxy that keeps answering normally while the game server behind it is down can look playable to us, so outages on proxied networks may be undercounted. And uptime says nothing about lag inside the game.

The definitions, ping timeouts and placeholder list are on the methodology page. How long listings keep running at all is a separate question, answered in how long Minecraft servers last.