Slow disk writing performance on Unraid

[yriche]

Hi! Not sure if this is the right place to ask about this since I’m not sure where in the stack this issue is coming from.

I have a NeoForge 21.1.209 server set up through the docker container. Recently we have been seeing really severe lag spikes and slow chunk loading

I took a spark profile of the major spikes (all of them were >1000ms) and saw that all of them were spending nearly 100% of the tick in PlayerList.saveAll(), stuck on a file write.

I tried quite a few things to troubleshoot, including testing the current world without mods, in a new world with and without mods, and nothing seems to help here.

The docker’s install folders are all on an unRAID share that’s set to cache only, so it should only be using the single NVMe drive in my cache pool. It was initially formatted to btrfs; I tried switching it to xfs when I saw this discussion but it didn’t seem to help.

The share is also exported and accessible via SMB, though I’m not sure if that affects write speeds

Has anyone seen a similar issue before?

[itzg]

This isn't the best place to ask since containers are nothing more than Linux processes. It's probably best to ask around with Unraid.

What would be helpful to share is where is the container's /data volume path mounted from? That will be the filesystem/disk to investigate with Unraid.

[yriche]

/data is mounted from a share that maps to an nvme drive on the server.

To be honest I’m not 100% sure it’s an Unraid issue…but I need to do a bit more troubleshooting to see if the issue persists on fresh installs/worlds. And I guess even then it might be a Neoforge or Minecraft issue.