Increased memory usage / out of memory errors

[timosmit]

After upgrading our Docker container to version 5.5.0, we noticed an increase in memory usage. This increase eventually resulted in several containers running out of memory. We have since reverted our image to version 5.4.0, which ceased the out of memory errors from happening.

I have attached several graphs of:

• the amount of requests that the container receives
• the CPU usage
• the RAM usage

We updated our containers ~ 21 January and reverted them ~ 27 January.

We run our instances based on the official maptiler/tilesever-gl image. They are capped at 4 Gigs of memory, which was seemingly enough for the past couple of years. According to our alerting channels, OOMs only occurred once over 2024 and 2025 with the current configuration (memory was last increased in 2023). Over the course of 21 January - 27 January I count 8 occurrences. If you require further contextual information, please let me know.

[alex-zyl]

We encountered the exact same problem. With 5.4.0 we could run replicas with 2Gi pod memory limit while memory utilization stayed at 1.8 max. After upgrading to 5.5.0, we had to raise the limit to 3Gi so it's at least no longer OOM-killed, but the utilization was still at ~2.9.

[okimiko]

I can confirm the increased memory consumption here, too.

After startup the newer version (5.6.0pre2 in my case) started a slightly lower (than 5.4.0), but after running a test script (generating 4096 PNG-Tiles) and waiting a bit to settle I can see 587MB (5.6) vs 385MB (5.4).

PS: As I was writing the amount of memory increased to 612MB for 5.6 without further requests (!) and was stable for 5.4.

At the moment I have no clue what is causing the increase, we have several candidates

• Ubuntu 22 -> 24 (with all its dependencies)
• Node 22 -> 24 (with all its dependencies)
• Tileserver + Dependencies
  • Maplibre-Native?

[utagawal]

Have you tried with the latest pre release ? Do you see the same behavior ?