# Server lag: choose checks for delays, freezes or low FPS “Lag” can describe a slow game loop, a delayed network connection or a client struggling to draw the scene. Choose the symptom before changing RAM or removing content. This route starts with observation and bounded measurements. It does not assume an overloaded host, a faulty plugin or insufficient memory from one graph. ## What happens during the problem? Follow a symptom link using mouse, touch, or Tab and Enter. These sections are ordinary visible HTML and do not run a diagnostic program. - [One player's picture stutters or FPS drops.](#bkmrk-w10-lag-fps) - [Players move back, actions arrive late or ping spikes.](#bkmrk-w10-lag-network) - [Everyone's game actions slow down together.](#bkmrk-w10-lag-ticks) - [Short freezes happen during travel or scheduled work.](#bkmrk-w10-lag-spikes) - [The game freezes, disconnects everyone or goes offline.](#bkmrk-w10-lag-offline) - [This is a game other than Minecraft.](#bkmrk-w10-lag-other) ## 1. Is the display slow for one player? **Check:** ask another player at the same location whether interaction timing is also slow. Have the affected player record their FPS and compare the scene without optional shaders or a local visual pack, preserving their settings first. Change one client option at a time. **If FPS improves while other players and game timing remain normal:** continue client graphics or pack troubleshooting. That result does not justify changing the server world. **If everyone sees slow interactions too:** [check server tick timing](#bkmrk-w10-lag-ticks). **If FPS is normal but actions arrive late:** [check the network pattern](#bkmrk-w10-lag-network). ## 2. Who experiences the delay? **Check:** note the time, affected players and region, whether voice or other applications also struggle, and whether server tick measurements stay normal. Compare a second authorised player's report during the same incident. Avoid publishing player addresses. **If one network is affected:** investigate its local connection with the player; a permitted alternate connection can provide comparison evidence. **If several independent networks are affected while ticks look normal:** [request a network-path investigation](#bkmrk-w10-lag-support). **If ticks also deteriorate:** [collect a server-side sample](#bkmrk-w10-lag-ticks). Healthy ticks do not by themselves prove the network caused the delay. ## 3. Does the server game loop slow down? For supported Paper installations, follow [the TPS/MSPT and spark guide](https://help.cloudexa-hosting.com/books/game-hosting/page/diagnose-minecraft-lag-with-tps-mspt-and-spark). [Paper's command reference](https://docs.papermc.io/paper/reference/commands/) identifies 20 TPS as the target and below 50 milliseconds per tick as the budget for that rate. These numbers are Minecraft tick metrics, not universal targets for every game. **If timing deteriorates during the complaint:** take one short, representative profile with the documented tool. Current Paper includes spark; older or other server platforms require their own supported setup. Follow [the official profiling instructions](https://docs.papermc.io/paper/profiling/), and share reports privately because metadata can reveal your server configuration. **If timing remains normal:** revisit [network evidence](#bkmrk-w10-lag-network) or [client FPS](#bkmrk-w10-lag-fps). **If the profile points to a component:** compare its documented behaviour and reproduce on a test copy before changing production. A sampled stack is a clue, not automatic proof that the named plugin is defective. ## 4. Do freezes match travel or scheduled jobs? **Check:** record the spike times alongside exploration, backups, map renders, world generation and plugin jobs. Check free storage and whether more than one heavy task overlaps. Correlation helps choose a next check; it does not establish the cause. **If spikes follow travel into new terrain:** review [planned terrain generation](https://help.cloudexa-hosting.com/books/game-hosting/page/pre-generate-minecraft-terrain-with-chunky-and-monitor-disk-space), including its storage requirements. Do not start a generation job during the active incident. **If spikes match backup or render schedules:** review [the task schedule](https://help.cloudexa-hosting.com/books/game-hosting/page/schedule-game-server-backups-and-check-retention) and profile one representative period; preserve recovery coverage rather than disabling all backups. **If disk writes fail or quota is exhausted:** use [the storage investigation](https://help.cloudexa-hosting.com/books/game-hosting/page/investigate-and-recover-from-a-full-game-server-disk). Never delete worlds or persistent plugin data as a performance experiment. **If no pattern emerges:** [send the time window and observations](#bkmrk-w10-lag-support). ## 5. Is this an exit or watchdog event? **Check:** preserve the first error before and after the freeze. Use [the Paper watchdog guide](https://help.cloudexa-hosting.com/books/game-hosting/page/read-a-paper-watchdog-report-without-disabling-its-protection) for a long-tick report, or [the memory-versus-process-kill guide](https://help.cloudexa-hosting.com/books/game-hosting/page/tell-a-java-memory-error-from-a-killed-minecraft-process) for an abrupt exit. The [Paper troubleshooting reference](https://docs.papermc.io/paper/basic-troubleshooting/) distinguishes severe stalls and exits requiring host evidence. Do not disable the watchdog, repeatedly force-kill the process or assume adding RAM solves a stall. If the console also stops responding or the whole service disappears, [ask support to inspect process and node evidence](#bkmrk-w10-lag-support). ## 6. Use the game's own measurements Use [Rust lag and entity checks](https://help.cloudexa-hosting.com/books/rust/page/diagnose-rust-server-lag-and-entity-load-without-wiping) or [FiveM resource performance checks](https://help.cloudexa-hosting.com/books/fivem/page/trace-fivem-script-errors-and-measure-resource-performance) where applicable. Do not apply Minecraft TPS thresholds or Java flags to these games. For another supported game, record its performance indicators and [ask for the appropriate route](#bkmrk-w10-lag-support). ## Outcome: compare one correction or escalate Keep a completed backup before changing configurations. Apply one documented correction and repeat the same observation under comparable normal play; undo it if behaviour worsens. Do not stage a load test against production. [Send support](https://help.cloudexa-hosting.com/books/getting-started/page/open-a-support-ticket) the service ID, build, times, affected players, relevant measurements, redacted first error and private profile link. Distinguish observations from guesses. [Return to symptom choices](#bkmrk-w10-lag-choices).