If a Minecraft server stops and the console shows only Killed, the operating system, container, or hosting panel probably terminated it after the process exceeded its memory limit. That is different from a Java heap failure, which normally prints OutOfMemoryError or another JVM error. Check the server and host logs first, identify the real memory ceiling, then set -Xms and -Xmx low enough to leave roughly 1,000–1,500 MB for Java overhead, the operating system, and other services. If the host cannot provide that headroom, reduce the workload or move to a larger memory limit.
Playing with friends? Godlike.Host runs hosted Minecraft Java servers from 10 data-center locations, with free DDoS protection.
What a plain Killed message means
The word Killed is not a Minecraft diagnosis. It is the shell or panel reporting that the Java process disappeared without printing a normal shutdown or Java exception. Several memory boundaries can be responsible:
As an Amazon Associate I earn from qualifying purchases.
- The Linux out-of-memory (OOM) killer: the operating system ended Java because the machine could not satisfy another allocation.
- A container limit: Docker, LXC, Kubernetes, or a similar container can enforce a memory ceiling below the physical RAM installed in the host.
- A hosting-panel limit: a game panel may stop the process when its assigned quota is reached, even when the underlying VPS has unused memory.
- A Java heap limit: Java may reject an allocation after reaching
-Xmx, usually leaving anOutOfMemoryErroror related JVM message rather than a bareKilled.
Paper’s troubleshooting guidance identifies a panel limit set too close to -Xmx as a common cause of out-of-memory kills. It also warns that failing to leave memory for the operating system or using an improper container limit can crash the server. Therefore, do not assume that the RAM advertised for a VPS is the amount available to the Minecraft process.
Java heap exhaustion versus an OS or panel kill
Use the final lines of every relevant log to classify the failure before changing flags. The distinction determines whether you lower the heap, raise a quota, or investigate a non-memory problem.
#1 Best Overall
- A-Tech 16GB RAM Module, DDR4 SO-DIMM 260-Pin, 3200MHz PC4-25600 (PC4-3200AA)
- Non-ECC Unbuffered, JEDEC DDR4 Standard 1.2V Operating Voltage
- Compatible with select Laptop, Notebook, Mini PC, and All-in-One (AIO) systems. Please verify your system's memory type, form factor, and maximum supported capacity before purchasing
- Not compatible with desktop DIMM, non DDR4 memory, or ECC memory types such as RDIMM, LRDIMM, and ECC UDIMM
- Increases available memory capacity to enhance system responsiveness, application performance, and multitasking capabilities.
Signs of a Java heap failure
- The server log contains
java.lang.OutOfMemoryError, “Java heap space,” “GC overhead limit exceeded,” or a similar JVM exception. - The Java process remains alive long enough to print a stack trace, crash report, or controlled shutdown.
- The configured
-Xmxvalue is reached while the host still has meaningful free memory and the container or panel quota has not been exceeded.
Signs of an operating-system, container, or panel kill
- The console ends with only
Killed, or the process vanishes without a Java exception. - The kernel log records an OOM event, or the panel reports that the container exceeded its memory quota.
- The host limit is equal to, or only slightly above,
-Xmx; Java’s non-heap memory and native allocations then consume the remaining allowance.
A plain termination is not proof that the Java heap was too large. It may mean the total process exceeded a hard external limit. Conversely, a Java exception does not mean the host has no spare RAM; it may simply mean -Xmx is too low for the configured workload. Always match the error to the limit that was actually reached.
How Minecraft server memory is divided
-Xmx is the maximum Java heap, not the maximum size of the complete server process. The process also needs memory for the Java runtime, thread stacks, class metadata, JIT-compiled code, direct and native buffers, memory-mapped files, plugins or mods, and the operating system itself. A world can therefore hit a container quota even when the heap has not reached its ceiling.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →-Xms is the initial heap size. Setting it equal to -Xmx reserves the whole heap immediately; that can make a small host run out of memory before the server has finished starting. A lower initial value can provide startup headroom, although changing only -Xms will not solve a workload that genuinely needs more heap.
Rank #2
- Simple design to perfectly protect the cooling module with high thermal conductive adhesive
- Supports Intel & AMD motherboards
- Selected high-quality IC
- Supports XMP2.0
- Energy saving with ultra-low working voltage
Paper’s Aikar flags guidance says not to allocate all available memory on a shared host and recommends reducing both -Xmx and -Xms by about 1,000–1,500 MB to avoid an OOM killer termination. Treat that range as a starting reserve, not a universal guarantee. Increase the reserve when the same machine also runs a database, proxy, panel, backup job, web service, or other game server.
Diagnose the real memory ceiling
- Stop repeatedly restarting the server. Repeated OOM kills can obscure the original log lines and create more load while the host is already constrained.
- Save the end of the Minecraft log. Inspect the last several dozen lines in the server’s log directory and note whether they contain an explicit Java exception or end abruptly.
- Check the host or container log. Ask the provider or inspect the panel’s event/history view for an “out of memory,” “limit exceeded,” or forced-stop event. On a Linux host, the administrator can check kernel messages for an OOM record; a common read-only check is
dmesgor the system journal, subject to the permissions of the account. - Read the enforced quota, not just the advertised plan. In a container or panel, record the hard memory limit shown for the server. On a self-managed Linux machine,
free -hshows host memory, while cgroup or container settings determine what Java can actually consume. If those numbers differ, use the lower enforced limit for planning. - Record the JVM command line. Confirm the active
-Xmsand-Xmxvalues in the startup script or panel’s Java options. Do not infer them from a hosting plan name. - Check co-hosted services. Account for databases, proxies, panels, scheduled backups, map renderers, voice services, and other processes that draw from the same limit. A Minecraft server is safe only when the sum of those processes and their growth has headroom.
- Reproduce cautiously. If the logs show no OOM evidence, test with the normal player and plugin workload while monitoring memory. A kill that occurs only when a particular plugin, world, or job runs points to a workload spike rather than a permanently undersized heap.
Set a safe heap budget
Start with the actual physical or plan memory, subtract memory required by the operating system and every co-hosted service, and then choose -Xmx below what remains. Paper’s documented reserve is approximately 1,000–1,500 MB below the available total when using its Aikar guidance; use more when the machine has other significant workloads.
| Situation | What to measure | Safer action | What is not established |
|---|---|---|---|
| Computer with 4 GB total RAM | Total system memory, desktop or background services, and the server’s hard limit | Minecraft Help says to decrease Minecraft allocation to 2 GB for stability | No player-count guarantee is published for that allocation |
| System with 8 GB or more | Usable memory after the operating system and other services, not the headline capacity | Minecraft Help says allocation may be increased, but choose it only after reserving non-heap headroom | No universal Xmx value or capacity benchmark is provided |
| Shared host or container | The panel or container hard limit and all processes charged to it | Keep roughly 1,000–1,500 MB below the available limit as a starting reserve, and more for co-hosted services | The provider’s advertised VPS RAM may not equal the server’s quota |
| Dedicated machine used only for Minecraft | Installed RAM, current system use, and expected workload spikes | Set heap below total RAM and retain operating-system and Java overhead; raise RAM if no safe reserve remains | A specific DIMM, SODIMM, or module capacity cannot be selected without hardware specifications |
For example, a 4 GB machine should not be given a 4 GB heap: Minecraft Help’s published example is a 2 GB allocation for stability. On a host with a larger limit, do not simply set -Xmx to the quota. Subtract the reserve first, then test under the heaviest routine workload. The correct value depends on world size, player activity, plugins or mods, and neighboring services, none of which can be reduced to one safe number.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Rank #3
- Requires overclocking/BIOS adjustments. Maximum speed and performance depends on system components, including motherboard and CPU.
- G.SKILL RipjawsV Series DDR4 U-DIMM Memory Kit, Model: F4-3200C16S-16GVK
- Non-ECC, DDR4 U-DIMM, 288-pin, for Desktop PC & Gaming
- Includes JEDEC default profile, and Intel XMP memory overclock profile
- Do not mix memory kits. Memory kits are sold in matched kits that are designed to run together as a set. Mixing memory kits will result in stability issues or system failure.
Repair sequence for a server that is being killed
1. Lower both heap values
Edit the panel’s Java options or the startup script so that -Xms and -Xmx fit below the enforced limit. Keep the same units consistently, for example 2048M or 2G. Do not leave an old, larger value in a second startup field; duplicate options can make it unclear which setting is active.
Use the Paper reserve as the initial calculation: reduce the two values by about 1,000–1,500 MB from the memory available to this server. If the host also runs a database, proxy, panel, or several services, reserve additional memory rather than consuming the entire remainder. Restart once, then inspect both the Minecraft log and the host or panel event log.
2. Correct the panel or container limit
If the panel quota is only a few megabytes above -Xmx, lowering Java alone may not be enough if the panel has a separate startup or process overhead allowance. Increase the container limit when the provider permits it, or set the heap lower so the total Java process fits comfortably. A panel limit is a hard boundary; free RAM elsewhere on the physical host does not bypass it.
Rank #4
- Simple design to perfectly protect the cooling module
- High thermal conductive adhesive
- Supports Intel & AMD motherboards
- Selected high-quality IC, Supports XMP2.0
- Lifetime warranty
3. Use Paper’s documented flag template carefully
When the server runs Paper, begin with the Aikar flag template documented by Paper and change only the memory values to match the budget you calculated. Paper cautions that anyone unsure what a JVM flag does should avoid using it. Do not copy flags from an unrelated version, startup script, or hosting panel without checking for duplicate options and compatibility. The flags are a configuration starting point, not a promise of a performance improvement or immunity from OOM kills.
4. Reduce workload when the heap is still exhausted
After the external limit and heap values are correct, reduce the amount of live work the server performs:
- Lower view distance so fewer chunks must remain active around players.
- Lower simulation distance to reduce ticking and entity updates outside the immediate play area.
- Stagger backups, map rendering, world pregeneration, and other concurrent jobs instead of running them during peak play.
- Disable or remove plugins and mods that are unnecessary, outdated, or associated with the memory spike; test one change at a time so the cause remains identifiable.
- Reduce simultaneous player or automation activity while you verify stability, then increase it gradually under monitoring.
These are operational tuning steps, not fixed values guaranteed by Paper or Minecraft Help. If the server remains stable only with an impractically small workload, the host does not have enough memory for the desired configuration.
Best Value
- Requires overclocking/BIOS adjustments. Maximum speed and performance depends on system components, including motherboard and CPU.
- G.SKILL RipjawsV Series DDR4 U-DIMM Memory Kit, Model: F4-3200C16D-16GVKB
- Non-ECC, DDR4 U-DIMM, 288-pin, for Desktop PC & Gaming
- Includes JEDEC default profile, and Intel XMP memory overclock profile
- Do not mix memory kits. Memory kits are sold in matched kits that are designed to run together as a set. Mixing memory kits will result in stability issues or system failure.
5. Re-test with evidence
Run the same activity that previously caused the termination: login bursts, normal exploration, redstone or mob farms, scheduled jobs, and plugin operations. Record the peak memory shown by the panel or operating-system tools, the active heap size, and the exact time of any failure. A successful restart is not proof of a fix; stability must include the workload that triggered the kill.
Recommended Free Tools
Common mistakes that make low-RAM crashes worse
- Setting
-Xmxequal to total RAM: this leaves nothing for Java’s non-heap allocations or the operating system. - Trusting the VPS headline: a plan may advertise total host RAM while the game container has a smaller hard quota.
- Increasing memory without checking the log: a bare
Killedcan be a panel or kernel action, while a Java exception may require a different fix. - Changing many settings at once: replacing flags, removing plugins, and changing distances simultaneously prevents you from identifying the effective repair.
- Copying an aggressive flag set blindly: Paper advises users who do not understand a JVM flag not to use it.
- Ignoring temporary jobs: backups, rendering, pregeneration, and updates can create short-lived peaks that exceed an otherwise adequate quota.
- Assuming a restart cures the cause: restarting releases memory temporarily but does not change the limit or workload that triggered the kill.
When to add RAM or move the server
Upgrade when you have already verified the quota, retained a sensible non-heap reserve, and reduced avoidable workload, but the desired player, world, plugin, or mod workload still exceeds the available budget. The direct remedies are compatible physical RAM or a host/container plan with a higher memory limit.
Buying compatible physical memory
Before ordering a module, verify the machine’s DDR generation, whether it accepts DIMM or SODIMM modules, the maximum supported capacity, slot count, and whether modules must be matched. The Minecraft allocation figures do not identify a compatible part; motherboard, laptop, or server specifications are required. After installation, confirm the operating system sees the new capacity before raising -Xmx.
Moving to a larger host or container
Compare the new service’s enforced container limit, not only its marketing total. Confirm that the limit covers Java overhead, the operating system or panel reserve, backups, and any proxy or database you plan to run alongside Minecraft. Migration also has operational costs: world transfer, configuration, permissions, plugin or mod compatibility, DNS or address changes, and a maintenance window. A larger quota is useful only if it is genuinely available to the process.
| Option | Best when | Advantages | Trade-offs |
|---|---|---|---|
| Lower heap and tune workload | The kill is occasional and the current limit can retain headroom | No hardware purchase or migration; changes are reversible | May reduce view distance, concurrency, or plugin/mod scope |
| Install compatible RAM | You control the hardware and the platform supports an upgrade | Raises capacity for Minecraft and other local services | Compatibility, installation, and maximum-capacity constraints must be verified |
| Move to a higher-memory container or host | The current quota is fixed or the workload outgrew the machine | Provides a larger enforced budget without changing local hardware | Migration work, recurring cost, and possible provider-specific limits |
Prevent the next OOM kill
- Write down total physical or plan memory and the actual container or panel limit.
- Keep
-Xmxbelow that limit and preserve approximately 1,000–1,500 MB as a starting reserve, increasing it for co-hosted services. - Keep
-Xmsfrom consuming the entire quota at startup. - Review the final server log and host or panel event after every crash.
- Monitor memory during peak players and scheduled jobs, not only while the server is idle.
- Test plugin, mod, view-distance, and simulation-distance changes individually.
- Document the active startup command so a panel update or script edit does not silently restore an unsafe heap.
- Upgrade the physical RAM or enforced host limit when the required workload cannot fit while preserving headroom.
The practical rule is simple: a Minecraft server needs memory outside the heap. A bare Killed message calls for evidence from the kernel, container, or panel as well as the Minecraft log. Once you know the real limit, lower -Xms and -Xmx enough to reserve non-heap and system memory, tune workload spikes, and increase the available limit when those changes cannot provide a stable margin.
Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallQuick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
