Paper Server Crashing or Not Starting: Find the Plugin Causing It
When a Paper server crashes, stops on startup, or lags badly, the cause is very often a plugin. Paper's troubleshooting documentation gives a method for finding it that works without any special tools: stop the server, read the logs, and narrow the list of plugins until the problem disappears.
This guide follows that page, checked on 2026-11-17, and adds panel steps. For a broader list of Minecraft startup problems see the won't-start fixes, and for lag measurement see the spark guide.
- Stop the server before changing files. Paper warns that modifying files while it runs will corrupt them. Back up first.
- Errors appear in the console, in
latest.login thelogsfolder, and incrash-reportsafter a crash. - If a plugin's name is in the stack trace, check that plugin first. Otherwise, disable plugins in halves to find the one responsible.
- A watchdog dump is not a crash. It means a tick took 10 seconds or more.
- Never force a newer world open with
-DPaper.ignoreWorldDataVersion=true. Paper says it may permanently corrupt the world.
On this page
Before you start
- Stop the server
Use Stop on the Console tab. Paper says modifying server files while it is running will corrupt them.
- Back up
Take a Whole server backup on the Backups tab and star it. See the backup guide.
- Find the error
Open the Files tab. Check
logs/latest.log, and thecrash-reportsfolder if the server crashed. Errors also show on the Console tab.
How to read a stack trace
A stack trace starts with an exception type and message, followed by lines beginning with at. According to Paper's documentation, the top at line shows where the problem began. Look in the whole trace for a plugin name. If you see one, start with that plugin: check for an update, read its documentation for misconfiguration, and contact its author if needed.
[Server thread/ERROR]: Could not pass event PlayerJoinEvent to ExamplePlugin v1.2
java.lang.NullPointerException: Cannot invoke ...
at com.example.exampleplugin.Listener.onJoin(Listener.java:42)The example above is made up to show the shape. Your trace will differ. Searching the exception type and message together is also a good next step.
Find the faulty plugin by splitting the list
Paper's recommended technique is a binary search. First confirm plugins are the cause by disabling all of them, for example by renaming the plugins folder, and starting the server. If the problem is gone, narrow it down:
- Split in two
Move half of the plugin jars into a folder outside
plugins, such asplugins-off. Keep plugins that depend on each other together in the same half. - Test
Start the server and try to reproduce the problem.
- Pick the half
If the problem is gone, it is in the half you removed. If it remains, it is in the half you left. Take that half and split it again.
- Repeat until one is left
Continue halving until a single plugin, or a single dependency pair, reproduces the problem.
The server does not start
- Check logs/latest.log first. If no logs are written at all, the start command is the problem. On ChicagoServers the launch line is staff-only, so contact support.
- Failed to bind to port: Paper says another server may be using the port, or
server-ipinserver.propertiesmay be wrong. That setting controls network interfaces, not your external IP address, so it normally stays empty. See the server.properties guide and the IP and port guide. - World from a newer version: downgrading is not supported. Use the latest Paper, or restore a backup made before the upgrade.
- Circular plugin loading: remove the offending plugin or report it. Paper mentions a legacy plugin loading flag as a last resort that may cause hard-to-debug issues.
- Outdated Java: Paper says to update Java and make sure it is on your PATH. On our panel the Java version is handled for you, so ask support if the log complains about it.
The server crashes or stops unexpectedly
- Keep the server updated and back up before each update. See the Paper update guide.
- Unexpected graceful shutdown: Paper says to enable
debuginserver.propertiesto capture a stack trace on the next shutdown. - Crash without logs: the process may have been killed for running out of memory. Paper's Linux advice is to reduce the maximum heap by 1 to 2 GB, or raise the container limit. On ChicagoServers, memory comes from your plan and the launch options are staff-only, so contact support rather than guessing. See how much RAM a Minecraft server needs.
Watchdog dumps and lag
A watchdog dump in the logs is not a crash. Paper says it appears when the server took 10 seconds or more to finish a tick. Read the first lines of the trace for the cause, and consider a spark profile. Paper gives /spark profiler start --timeout 300 as an example, and adds --only-ticks-over when you are chasing spikes.
- High RAM use is usually expected, because the heap grows and rarely shrinks. Out-of-memory crashes come from too little system memory or container limits.
- The main tick loop mainly uses one thread, so more cores do not always help. Paper suggests at least 4 threads available.
When to ask for help
Paper points to its community Discord for support, GitHub issues for bugs, and the plugin author for plugin problems. Have the exception type, the top at line, your Paper version and the plugin list ready. If you are stuck, contact support.
Frequently asked questions
How do I find which plugin is crashing my Paper server?
Stop the server, back up, then check whether the stack trace names a plugin. If it does not, disable all plugins to confirm they are the cause, then use a binary search: disable half, test, and keep halving. Keep dependent plugins in the same group.
What is a watchdog dump?
Paper says it is not a crash. It appears when the server took 10 seconds or more to finish a tick. Read the top of the trace for the cause and consider a spark profile.
Where are Paper's error logs?
In the console, in logs/latest.log, and in the crash-reports folder after a crash. If no logs are written at all, the problem is probably the start command.
Why does my server say failed to bind to port?
Paper lists two causes: another server is already using the port, or server-ip in server.properties is set wrongly. That setting selects network interfaces, not your external IP, so it normally stays empty.
Is it safe to force the server to start on a newer world?
No. Paper says downgrading is not supported and that forcing startup with the world data version flag may permanently corrupt the world. Use the latest Paper or restore a backup.