Lodehost
Server lag or connection lag?
A Minecraft server runs its world in ticks, twenty per second when all is well, which leaves 50 milliseconds for each tick. Server lag means ticks take longer than that, and everyone online feels it at the same moment. Connection lag lives on one player's route to the server, and only that player feels it.
| What players notice | Most likely | Check with |
|---|---|---|
| Everyone at once: mobs freeze, broken blocks come back, crops stop | Server lag | spark tps |
| One player rubber-bands while the rest play fine | That player's connection | spark ping |
| The whole server stalls for seconds, every few minutes | A lag spike or memory pauses | spark gc and a spike profile |
| Only while someone explores new land | Terrain generation | The pre-generation guide |
Read the tick times first
Console
spark tpsIn the console the command goes without a slash; in the game you type /spark tps and need operator rights or the spark permission. The answer shows the ticks per second over several recent periods, how long ticks took, and how busy the processor was.
| Tick duration | What it means |
|---|---|
| Clearly below 50 ms | Room to spare. The server keeps up with 20 ticks per second. |
| Close to 50 ms | At the limit. One busy moment tips it over. |
| Above 50 ms | The server falls behind and the tick rate drops below 20. |
| A normal middle value with a very high maximum | Spikes: most ticks are fine and a few are very slow. |
The tick rate alone hides a lot, because it never goes above 20. A server at 20 ticks per second whose ticks take 48 milliseconds is one farm away from lag. The durations show that; the tick rate does not.
Profile the lag while it happens
Wait for the lag
A profile of a quiet server shows nothing useful. Start it on a busy evening, or while players stand near the build that seems to cause the trouble.
Start the profiler
Type
spark profiler start --timeout 300in the console. spark samples what the server is doing for five minutes and then stops by itself.Keep playing as usual
Do not restart or change settings while it records, or the profile mixes two different situations.
Open the report
When the profile ends, spark prints a link to a viewer page. Open it in a browser; there is nothing to install.
Console
spark profiler start --timeout 300
spark profiler info
spark profiler stopspark profiler info tells you whether a profile is still running. spark profiler stop ends it early and prints the link, and spark profiler cancel throws the recording away without a report.
Catch spikes instead of averages
When the server runs fine most of the time and freezes now and then, an ordinary profile averages the freezes away. Record only the slow ticks instead:
Console
spark profiler start --only-ticks-over 100 --timeout 600This keeps only ticks that took longer than 100 milliseconds, twice the normal budget, for ten minutes. Whatever shows up in that report is what the freezes have in common.
For stalls of several seconds, run spark gc as well. It lists recent garbage collections and how long each one paused the server. Long pauses there point at memory pressure rather than a heavy plugin, and the RAM guide explains what to do about that.
Reading the report
- Start at the server thread. Open
Server threadand keep opening the child with the highest percentage. - Look for names that are not the game. A plugin appears under its own package name, deep in the tree. The Sources view adds up the time per plugin or mod, which is often the quickest answer.
- Entities. A large share spent ticking entities means too many mobs or items in loaded chunks: farms, villager halls, item piles.
- Block entities. Hoppers, furnaces and other working blocks in large numbers show up here.
- Chunk loading and generation. Players exploring new land or flying fast with an elytra.
Share the link, not a screenshot
The viewer link holds the whole tree. When you ask a plugin author or our team for help, send that link; a screenshot only shows the branch you happened to open.
A quick health check
Console
spark health
spark heapsummaryspark health puts the main numbers on one screen: tick rate, tick durations, processor and memory use. It is the first thing to run when someone reports lag and you only have a minute.
spark heapsummary goes a step further for memory and lists which kinds of objects fill it. Counting them pauses the server for a moment, so run it when few players are online.
What the usual culprits look like
| Found in the report | Usual cause | What to do |
|---|---|---|
| Entity ticking around one base | A mob farm or villager hall that kept growing | Shrink it, or lower the spawn limits in bukkit.yml |
| Hoppers and moving items | Long hopper chains and item sorters | Fewer hoppers, or water streams for transport |
| One plugin's package | That plugin, often a task on a timer or on every player move | Update it, check its settings, or replace it |
| Chunk generation | Players exploring new land | Pre-generate the world and set a world border |
Long pauses in spark gc | Memory pressure | Check heap use against the size of your plan |
A lower view and simulation distance helps against almost all of these, because it shrinks the area the server keeps alive around every player.
spark on other server types
Paper and Purpur include spark from 1.21 on. On Fabric, Forge and NeoForge it is a mod you add to the server, and the commands are the same. A vanilla server cannot load it; to profile a vanilla world, run the same world on Paper for an evening and measure there.
Written and checked by the Lodehost team, last reviewed on 14 September 2026.
