Hi Creators,
Today, we’re excited to launch server memory snapshots to help you identify the root cause of crashes and improve your experience’s stability. You can find the new page in Creator Dashboard for your experiences under Monitoring > Crashes.
Here’s what’s new:
Server Crashes
We added a new crash count chart to give you visibility into your server crashes. Game servers crash for two primary reasons:
- Out-of-Memory crashes: These are errors you can take direct action to stop. A spike in this line means servers are crashing due to high memory usage in your experience. This post explains how to identify and diagnose these root causes using the memory snapshots and provides links to more information on how to further debug and fix them in Studio.
- Platform crashes: These crashes are much less frequent, and unfortunately, they’re rarely directly actionable by a creator. They occur when the server enters an unrecoverable “bad” state, often due to engine-level errors. We work hard to monitor and automatically fix these errors platform-wide, but if you notice a significant uptick in these crashes, please help us out by filing a bug report so our team can investigate and resolve the issue faster.
Data is available for the most recent 7 days for games with 100+ DAU. If you see “No data for selected period”, that means you had 0 server crashes in that date range.
Server Out-of-Memory Snapshots
When a server detects it is running out of memory, the engine automatically captures a compact summary of the DataModel before the server shuts down. This summary lets you see where the memory is going so that you can identify and diagnose unexpectedly-large memory consumers.
Example Workflow: From Crash to Fix
Here’s a quick example of how you can use the above dashboard to find and fix an acute memory spike:
1. Identify the spike: You notice Out-of-Memory crashes spiking on your server crashes chart. You filter by Place version to confirm the crashes started immediately after your recent update.
2. Select a snapshot: You scroll down to the memory snapshots and examine the list by latest timestamp or ‘Server uptime.’ You select a snapshot from a server that crashed within minutes of starting up, which suggests a massive, acute memory spike during initialization.
3. Visualize the memory: The treemap shows nodes weighted by memory usage. You click into each node to see which services, folders, or assets dominated memory at crash time. Opening the snapshot in the treemap viewer highlights Clouds as the largest node. Clicking into Clouds shows that ParticleEmitter makes up 50% of the Clouds node. This indicates that heavy assets are being cloned into memory at startup but not properly parented or destroyed.
4. Investigate and fix: You use the viewer’s breadcrumbs to trace the context of the bloated node. Since certain assets are dynamically generated at runtime, you use this path as a starting point to locate the VFX system in Studio’s Explorer, add cleanup logic, and monitor your next update to verify the fix.
You can also download the raw CSV summary of the memory snapshots for your own analysis, including feeding the data into your favorite LLM or writing your own parsing scripts.
We hope these changes give you a better understanding of your server performance. Please let us know in the comments if you have any feedback or questions.
FAQs
What about client memory snapshots?
- We hear you! We are actively exploring similar diagnostic tools for client-side memory but prioritized server memory first to help you address widespread stability and session disconnect issues.
Are platform crashes my fault or Roblox’s?
- These are Roblox’s engine-level bugs where the server intentionally crashes to capture a report. If you see a sharp, sudden spike or a high sustained volume of runtime errors, file a bug report.
Where can I learn more about managing memory and reducing OOM crashes?
- Check out our CreatorHub documentation on Memory usage and Performance optimization and DevForum guide on Real world building and scripting optimization for Roblox for best practices.
Do high server-side crashes mean high client-side crashes too?
- Not necessarily. Server Out-of-Memory crashes happen when the game server exceeds its memory limits, disconnecting players from the session. Client crashes happen locally on a player’s device, though heavy assets can negatively impact both.
(chars chars chars)
