Our experience uses a lightweight lobby/start place to teleport players to several places within the same universe.
The destination places are substantially heavier, with large maps, and a high player count per server (e.g. 100+ players). However, they operated without this sudden pattern of crashes, including servers with much higher player counts. Recently, servers across both the lightweight Hub and multiple heavier destination places have begun crashing repeatedly. Most are classified as Out of Memory, while others in the same incident windows are classified as Platform Crashes.
The hub/start game can be found here. (Place ID 130236446350677)
Experience ID: 10111700956
As an example, a hub/start place server running from 11:40 AM to 12:47 PM BST crashed at 18,622MB used, and several replacement servers (around 4 more) also crashed with similarly high memory usage. There was then a crash with around 534 MB usage classified as a Platform Crash. The memory snapshots only show 48-71 MiB and none show the indicated high memory readings from the server management details so there is a large amount of unaccounted memory use in the snapshots.
The server crash logs show no consistency or relevance to the crashes. Initial crash usually occurs with error code 277 for players, then shows 279 when they attempt to rejoin.
Crash log examples, with 47GB memory for a 6 minute server:
Graph showing sudden spike in platform and OOM crashes:
We have been running similar games and setups for many years without issue.
Expected behavior
Public servers should normally remain fairly stable and did not start crashing until around ~2 weeks ago.
When an OOM occurs, the snapshot doesn’t account for the indicated memory used, and the server management filter does not categorise any of the crashes as OOM, only as a regular crash event.
A private message is associated with this bug report





