---

name: minecraft-server-admin description: "Set up, operate, tune, and troubleshoot Minecraft Java 26.x and legacy 1.21.x servers across Paper, Purpur, Folia, Velocity, Fabric, and NeoForge. Use for infrastructure, backups, proxies, and live operations, not plugin or mod development."

Minecraft Server Administration Skill

Scope and Routing Boundaries

Routing Boundaries

• Use when: the task is infrastructure or live operations for Minecraft servers (deployment choice, tuning, backups, proxying, security, incident response).
• Do not use when: the task is writing plugin code (minecraft-plugin-dev) or writing mods/loaders (minecraft-modding, minecraft-multiloader).
• Do not use when: the task is WorldEdit command workflows (minecraft-worldedit-ops) or EssentialsX workflow/policy design (minecraft-essentials-ops).
• Do not use when: the task is datapack/resource-pack authoring (minecraft-datapack, minecraft-resource-pack).

References

• Read references/deployment-checklists.md when the task is an incident, rollout window, proxy change, or recovery drill and you need a compact checklist before acting.

---

Deployment Decision Matrix

Use this table when choosing a deployment. For an existing server, inspect its stack and version first and preserve them unless migration is requested.

Deployment profile	Recommended stack	Pick this when	Watch-outs
Small SMP on Paper	Paper only	Up to ~30 concurrent players, broad plugin compatibility, low ops overhead	Do not over-tune early; profile before changing many defaults
Larger public Paper server	Paper + Spark + stricter plugin/change control	Public server with frequent joins, moderate plugin set, uptime matters	Plugin sprawl is the top TPS risk
Velocity-backed network	Velocity + multiple Paper backends	Hub/minigame/factions split across servers, need shared entrypoint	Forwarding/auth mismatches can block joins
Purpur gameplay-heavy server	Purpur (optionally behind Velocity)	You want gameplay knobs exposed in config without custom plugin code	Extra toggles increase misconfiguration risk
Folia high-concurrency server	Folia + Folia-compatible plugins only	Very high concurrency with region-threading goals	Many plugins are not Folia-safe
Fabric/NeoForge mod server	Fabric or NeoForge server build	You require loader mods, custom content, modpack behavior	Bukkit/Paper plugins do not apply

Deployment Type Routing

• Small SMP and most public plugin servers: use Paper baseline first.
• Use Purpur only when you explicitly need Purpur gameplay controls.
• Use Folia only when plugin compatibility has been validated for region-threading.
• Use Velocity when one process is not enough or you need separate backend roles.
• Use Fabric/NeoForge when the requirement is mod-driven, not plugin-driven.

Java versions

• Paper and Purpur: Minecraft 26.1+ requires Java 25; Minecraft 1.21.x uses Java 21. Verify the exact Paper/Purpur build and installed plugins before a version change.
• Current Velocity 4.x requires Java 25. For a legacy proxy, retain the Java version required by that exact proxy release rather than applying the current version by default.

---

Playbook: Performance Tuning

Step 1: Establish baseline

Collect a baseline before edits. Linux host example:

free -h
top -b -n 1 | head -n 25

Windows PowerShell host example:

Get-CimInstance Win32_OperatingSystem | Select-Object FreePhysicalMemory,TotalVisibleMemorySize
Get-Process java | Sort-Object CPU -Descending | Select-Object -First 5 Id,CPU,WorkingSet,Path

In server console:

tps
mspt
spark healthreport

Record:

• peak player count
• MSPT at idle and at peak activity
• top plugins by CPU from Spark report

Step 2: Profile the real bottleneck

Use Spark instead of guesswork:

spark profiler --timeout 180
spark tps
spark tickmonitor --threshold-tick 50

The 50 ms threshold reports ticks that exceed one normal 20 TPS tick; adjust it only when the incident threshold is intentionally different.

Then identify whether the issue is:

• plugin task load
• entity count / mob farm pressure
• chunk generation / disk I/O
• garbage-collection pauses

Step 3: Apply targeted fixes

Do not tune everything at once. Apply one group at a time.

1. Chunk and simulation pressure:

• lower view-distance and simulation-distance first
• pre-generate worlds for survival-heavy maps

1. Entity pressure:

• adjust despawn ranges and mob-spawn behavior in Paper world config
• cap problematic entities where appropriate

1. Plugin pressure:

• disable or replace top offenders from Spark traces
• reduce async task frequency where plugin settings allow

Step 4: Use stable startup flags

For Java 25 on current Paper/Purpur, start with a simple measured baseline:

java -Xms4G -Xmx4G -jar server.jar --nogui

Increase heap only when profiling and player load justify it. Add collector flags only after validating them against the exact JDK and Paper version.

Step 5: Verify improvements

After each change set:

spark tps
spark healthreport

Keep the change only if MSPT/tick stability improves under representative load.

---

Playbook: Plugin Operations

Safe plugin change workflow

1. Build a plugin inventory:

• plugin name and version
• required dependencies
• minimum server version

1. Stage updates:

• apply plugin updates in staging first
• run smoke checks: joins, teleports, economy, permissions, saves

1. Production rollout window:

• announce maintenance window
• stop server cleanly
• snapshot plugin jars and config
• deploy update batch

1. Post-rollout verification:

• watch startup logs for API warnings
• run command and permission sanity checks
• track TPS/MSPT for 15-30 minutes

Rollback checklist

• Keep previous plugin directory snapshot.
• If severe regression occurs:
  • stop server
  • restore prior plugin jars/configs
  • start and confirm login/world integrity
• Document failing plugin/version pair for future blocks.

---

Playbook: Proxy and Forwarding (Velocity)

Velocity baseline

velocity.toml essentials:

bind = "0.0.0.0:25565"
online-mode = true
player-info-forwarding-mode = "modern"
forwarding-secret-file = "forwarding.secret"

Backend server requirements

server.properties on backend:

online-mode=false

Paper backend forwarding support (config/paper-global.yml):

proxies:
  velocity:
    enabled: true
    online-mode: true
    secret: "paste-the-shared-forwarding-secret-here"

Use the same secret value stored in Velocity's forwarding.secret file; the backend config takes the secret string itself, not a file path.

Keep enforce-secure-profile at the server default unless you are handling a specific legacy-client or incident workaround. It is not part of the baseline Velocity setup. Also confirm settings.bungeecord: false in spigot.yml; do not enable BungeeCord forwarding and Velocity modern forwarding at the same time.

Validation runbook

1. Confirm proxy port is reachable from clients.
2. Confirm backend is firewalled from direct internet access.
3. Join through proxy and verify:

• UUID consistency
• skin/profile forwarding
• plugin permission behavior

Common proxy incidents

• "Invalid player info forwarding":
  • mismatch in forwarding mode or shared secret between Velocity and backend.
• Random auth/session failures:
  • mixed online-mode expectations or direct backend exposure.

---

Playbook: Backup and Recovery

Backup policy template

Asset	Frequency	Retention	Notes
Every configured world folder	Hourly incremental + daily full	7 daily, 4 weekly	Include custom and externally stored worlds
plugins/, config/, and root server state	Daily	14 daily	Required for operational restore
Proxy config/secrets	Daily	30 daily	Store encrypted off-host
Container/orchestration files	On change + weekly	8 weeks	Git-tracked where possible

Example backup script (Paper/Purpur)

For a production server, quiesce world writes before copying live world folders. The example below assumes a maintenance window and a cleanly stopped server. If you use RCON-based live backups instead, choose a client/secret mechanism that does not expose the password in command arguments. A safe implementation must run save-off, then save-all, copy the data, and guarantee save-on cleanup even when the copy fails. Do not treat a flush or save-all alone as a live backup protocol. Folia disables save-all; use a clean stop or a verified platform snapshot there. Test the restore path before trusting any backup.

For Fabric or NeoForge recovery, additionally inventory mods/ and record the exact loader launch state: Minecraft and loader versions, Java version, launch command and arguments, launcher or installer artifacts, and any version or library manifests used by that deployment. Restore those matching components with the configured world folders; this Paper/Purpur example does not capture them for you.

Supply the complete world-folder list as arguments, relative to SERVER_ROOT, after checking level-name and any multi-world plugin configuration. Back up worlds stored outside that root separately; do not infer coverage from world*.

#!/usr/bin/env bash
set -euo pipefail

if [[ "${SERVER_STOPPED_CONFIRMED:-}" != "1" ]]; then
  echo "Set SERVER_STOPPED_CONFIRMED=1 only after stopping the server cleanly." >&2
  exit 1
fi

DATE="$(date +%Y-%m-%d_%H-%M-%S)"
BACKUP_ROOT="/backups/minecraft"
SERVER_ROOT="/srv/minecraft"
DEST="${BACKUP_ROOT}/${DATE}"

if [[ "$#" -eq 0 ]]; then
  echo "Usage: backup.sh <world-folder> [additional-world-folders...]" >&2
  exit 1
fi
for world_dir in "$@"; do
  if [[ ! -d "$SERVER_ROOT/$world_dir" ]]; then
    echo "Missing configured world folder: $world_dir" >&2
    exit 1
  fi
done

mkdir -p "$DEST"

tar -czf "${DEST}/worlds.tar.gz" -C "$SERVER_ROOT" -- "$@"

state_items=()
for item in \
  plugins config server.properties bukkit.yml spigot.yml paper-global.yml \
  paper-world-defaults.yml permissions.yml ops.json whitelist.json \
  banned-players.json banned-ips.json; do
  [[ -e "$SERVER_ROOT/$item" ]] && state_items+=("$item")
done

if [[ "${#state_items[@]}" -gt 0 ]]; then
  tar -czf "${DEST}/server-state.tar.gz" -C "$SERVER_ROOT" "${state_items[@]}"
fi

Recovery drill (must be tested)

1. Stop server.
2. Restore selected backup to staging directory.
3. Validate ownership/permissions.
4. Start server in maintenance mode.
5. Verify:

• world spawn loads
• player data is readable
• key plugins initialize

1. Reopen to players.

Define targets:

• RPO (acceptable data loss window)
• RTO (acceptable restore duration)

---

Playbook: Live Troubleshooting

Incident triage flow

1. Classify incident:

• crash on startup
• severe lag / TPS collapse
• join/auth failures
• memory/disk pressure

1. Capture evidence first. Linux host example:

df -h
free -h

Windows PowerShell host example:

Get-PSDrive -PSProvider FileSystem
Get-CimInstance Win32_OperatingSystem | Select-Object FreePhysicalMemory,TotalVisibleMemorySize

From server/proxy console:

spark tps
spark healthreport

1. Stabilize:

• stop risky rollout changes
• disable newest suspect plugin first
• reduce player impact (maintenance mode, temporary queue, restricted worlds)

1. Recover and document:

• apply rollback or hotfix
• document root cause and permanent preventive action

Fast symptom map

Symptom	First checks	Typical root cause
Startup crash after plugin update	latest logs, plugin dependency chain	incompatible plugin/API mismatch
High MSPT only at peak hours	Spark profiler, entity/chunk stats	mob farms, heavy scheduled tasks, chunk I/O
Players cannot join via proxy	forwarding mode + secret + backend exposure	Velocity/backend config mismatch
Periodic hard lag spikes	GC pauses + autosave + backup overlap	memory pressure, backup I/O contention

---

Operational Config Reference

server.properties high-impact keys

max-players=100
view-distance=10
simulation-distance=8
sync-chunk-writes=true
max-tick-time=60000
enable-rcon=false

bukkit.yml and spigot.yml

Use Spigot/Bukkit docs as authoritative for version-specific defaults and side effects. Tune conservatively and re-measure after every change set.

Paper config files

• config/paper-global.yml
• config/paper-world-defaults.yml

Adjust these only after profiling identifies an actionable bottleneck.

---

Deployment Patterns

Docker Compose (Paper)

services:
  paper:
    image: itzg/minecraft-server:java25
    container_name: mc-paper
    environment:
      EULA: "TRUE"
      TYPE: "PAPER"
      VERSION: "26.3"
      MEMORY: "10G"
    ports:
      - "25565:25565"
    volumes:
      - ./data:/data
    restart: unless-stopped

This Docker example targets current Paper 26.3 and Java 25. Re-check plugin compatibility and take a restorable backup before changing an existing server's Minecraft or Java line.

Pterodactyl/Wings notes

• Keep startup command and memory allocations consistent with tested JVM flags.
• Pin panel and Wings versions to supported combinations.
• Validate backup mount and restore workflow before production.

---

Security Hardening Checklist

• Keep online-mode=true unless proxy forwarding requires backend online-mode=false.
• Never expose backend Paper ports directly when using Velocity.
• Restrict RCON and panel/admin surfaces by firewall/IP allowlist.
• Rotate secrets (forwarding.secret, panel credentials, backup credentials).
• Keep plugin sources trusted; verify checksum/release source before install.
• Maintain off-host encrypted backups.
• Keep Java runtime and host OS patched.

---

References

Paper (primary)

• https://docs.papermc.io/paper/
• https://docs.papermc.io/paper/getting-started/
• https://docs.papermc.io/paper/admin
• https://docs.papermc.io/paper/profiling/

Spigot/Bukkit (primary config reference)

• https://www.spigotmc.org/wiki/spigot-configuration/
• https://www.spigotmc.org/wiki/bukkit-configuration/
• https://www.spigotmc.org/wiki/start-up-parameters/

Proxy and ecosystem docs

• https://docs.papermc.io/velocity/
• https://docs.papermc.io/velocity/getting-started/
• https://docs.papermc.io/folia/faq/
• https://spark.lucko.me/docs
• https://spark.lucko.me/docs/Command-Usage
• https://github.com/itzg/docker-mc-backup
• https://geysermc.org/wiki/geyser/
• https://pterodactyl.io/project/introduction.html