Skip to content
←Return to Projects
Case Study

IceLegends SMP

A vanilla-first Romanian Minecraft SMP built for clean 3–4 month seasons.

The Problem

Running a community SMP means balancing plugin depth against vanilla feel, managing a world-reset cycle, and keeping operations simple enough for a solo admin. IceLegends SMP solves this with 11 carefully chosen plugins, strict feature gating per a "vanilla-first" principle, and fully automated backups and world resets.

My Role

Designed, configured, and deployed solo — server architecture, 11-plugin selection and configuration, LuckPerms permission model, economy and nation design, Docker infrastructure, and Cloudflare Tunnel ingress for Dynmap.

Join the Server ↗

Highlights

  • PaperMC 1.21 on Docker with a hard 11-plugin cap — every plugin is justified and none are redundant
  • Towny nations with seasonal quests (Quests + Citizens NPCs) and a single manually-spawned MythicMobs boss per season
  • AureliumSkills passive bonuses only — active abilities disabled to preserve vanilla combat (P-01)
  • DiscordSRV bridge logs every death, join, and chat event to dedicated Discord channels in real time
  • Daily automated backups with 7-day retention; resource world resets every Monday at 05:00 via cron

Stack

PaperMC 1.21DockerCloudflare TunnelLuckPermsTownyAureliumSkillsMythicMobsDiscordSRVDynmap

Constraints

  • Solo admin — every operation must complete in under 10 minutes (P-02).
  • Vanilla-first (P-01): if any plugin disappeared tomorrow, the game must continue unchanged.
  • 11-plugin hard cap — every addition must displace something or prove unique, irreplaceable value.

System Architecture

Edge
Cloudflare Tunnel (Dynmap web map)
TCP :25565 open (Minecraft clients)
Runtime
Docker + docker-compose
itzg/minecraft-server image
Core
PaperMC 1.21
Vault
LuckPerms
EssentialsX
CoreProtect
Gameplay
Towny
AureliumSkills
ShopGUI+
Quests + Citizens
MythicMobs
Multiverse-Core
Ops
DiscordSRV bridge
Daily backup cron (04:00)
Weekly world reset cron (Mon 05:00)

Key Trade-offs

The decisions worth defending — what I chose, what I turned down, and why.

Plugin count cap

Chose

Hard cap at 11 plugins

Rejected

Unlimited plugin stack

Each plugin is a maintenance surface and a potential conflict point. The cap forces prioritisation and keeps the dependency graph legible for a solo admin.

Public ingress for Dynmap

Chose

Cloudflare Tunnel (no open ports)

Rejected

Expose :8123 directly via UFW

No origin IP exposure, free DDoS protection, and zero firewall rule maintenance — the same pattern used by the IceLegends portfolio site.

AureliumSkills scope

Chose

Passive bonuses only, active abilities disabled

Rejected

Full AureliumSkills feature set including active combat abilities

Active abilities alter vanilla combat — a direct violation of P-01. Passive bonuses reward playtime additively without changing the combat model.

Season lifecycle

Chose

Full world reset every 3–4 months

Rejected

Incremental map expansion with no reset

Resets keep the world fresh, prevent terrain exhaustion, and give every new player an equal start — critical for community health in a small SMP.

What I'd Do Differently

An honest retrospective — the stuff I'd change with more time, more users, or a second pass.

  1. 1Add a staging compose stack so plugin updates can be tested before touching the live season — one bad update cost several hours of rollback work.
  2. 2Move backup storage off-site from day one; local-only archives are a single hardware failure away from a full season loss.
  3. 3Wire Dynmap as a named Docker service rather than relying on the host port, so the Cloudflare Tunnel config survives container IP reassignment on restart.

Technical Deep-Dive

Architecture, specifications, and implementation details.

17 — Spark — Performance Profiler

"

Diagnostichează lag-ul în timp real — tool nr. 1 pentru orice admin Paper

#De ce e necesar

Fără Spark, când TPS-ul scade sub 18 nu știi DE CE. Cu Spark, dai o comandă și în 60 de secunde ai un raport vizual care îți arată exact:

  • Ce plugin consumă cel mai mult CPU
  • Ce entități sau chunk-uri cauzează lag
  • Unde e bottleneck-ul în tick loop

#Download

Sursă:   modrinth.com/plugin/spark
Licență: LGPL-3.0 — gratis
Impact vanilla: zero

#Instalare

# Copiază JAR în plugins/
cp spark-*.jar data/plugins/

# Restart server
docker compose restart mc

# Verifică că a pornit
docker compose logs mc | grep -i spark

#Comenzi esențiale

# Pornește profiling 60 secunde și generează raport
/spark profiler --timeout 60

# Raport instant de TPS și health
/spark health

# Monitorizare live în consolă
/spark tps

# Raport memorie (heap dump)
/spark heapdump

#Workflow când TPS scade

1. TPS < 18 detectat via /spark tps
2. /spark profiler --timeout 60
3. La final → link către raport web (spark.lucko.me)
4. Raportul arată top consumatori CPU în %
5. Identifici pluginul / entitatea vinovată
6. Acționezi: reduci view-distance, elimini entități, dezactivezi plugin

#Permisiuni LuckPerms

# Doar Staff poate accesa Spark
/lp group moderator permission set spark.* true
/lp group admin permission set spark.* true
# Jucătorii nu au acces

#Integrare cu 13_performance.md

Când TPS < 15 (Critical threshold din spec):

# Pas 1 — identifică cauza
/spark profiler --timeout 30

# Pas 2 — acționează pe baza raportului
# (nu restarturi la întâmplare)

#Related Documents

  • Performance targets → 13_performance.md
  • Runbook TPS → 14_backup_recovery.md
  • Plugin stack → 02_plugin_architecture.md
~ End of Document ~
Kiwi