Docker version causes high CPU load

[psychofaktory]

For some time now, this version of Docker has been causing extremely high CPU usage.
The load occurs even when no map is being updated.
As far as i can tell, this issue has existed since the introduction of the continuous map updates.
There are no anomalies in the log files, so I can't figure out what might be causing this.
Other than that, BlueMap works as expected.

BlueMap Version: bluemap-5.20-cli
Storage is Postgres SQL

[TBlueF]

The load occurs even when no map is being updated.

As far as i can tell, this issue has existed since the introduction of the continuous map updates.

I feel like these kind of contradict each other.. ? :D

BlueMap is using a lot of CPU while it is updating/rendering a map. This is expected behavior.
You can control how much cpu/cores it uses by changing the render-thread-count in the core.conf.
If bluemap is done updating the map and there are no further changes to the world-files it should go idle and barely use any CPU. But if your world is actively being played on, this probably happens very rarely. Even if no players are online, your server might still have some chunks loaded and produce changes.

With the information you provided i can not reproduce any high cpu usage while idle. As you can see here, the CPU usage drops to 0-1% as soon as the updating is done:

If you still think your issue exists, then please provide some more information that i can use to reproduce/analyze this :)

[psychofaktory]

I can't confirm that.

For example, there wasn't a single player online yesterday. Overnight, both the Minecraft server and the BlueMap container were restarted.
According to the logs, no map update was performed. Nevertheless, the container was using about 16% of my CPU (AMD Ryzen 7 PRO 5755G; 8C/16T).
I have since adjusted the render mask, thereby forcing a map update. The container is now also using about 16% of the CPU during the map update.
A few months ago, the idle CPU usage was still at the 0-1% you mentioned.

“render-thread-count” has always been set to “2” and has never changed.

I’m willing to provide further information. What additional details would be helpful for further analysis of this issue?

[TBlueF]

Useful information would be:

• Logs with timestamps
• A cpu-graph with timestamps
• A thread-dump during such a "idle but high cpu" moment .. (e.g. using jstack: docker exec -it <containerName> jstack > someFile.txt)

(The thread-dump being the most important info)

[TBlueF]

[...] the BlueMap container were restarted. According to the logs, no map update was performed. Nevertheless, the container was using about 16%

If the bluemap container just got restarted then it will always do at least a quick full map-scan, which looks just like a map-update in the logs. So i am not sure how there can be "no map-update" in the logs after a container-restart 🤔

The container is now also using about 16% of the CPU during the map update.

During a map update this is normal and expected.
You have 2 render-threads and 16 cpu-cores .. 16% of 16 cores is ~2.5 .. so it's using 2.5 cores .. 2 render-threads + 0.5 for some overhead and webserver. Completely normal :)

The only thing that would not be normal is if the cpu-usage wouldn't drop when everything is updated and there is nothing rendering.

---

Currently the information you are providing is a bit confusing :D .. So i will wait for the additional information that i listed above and we will be able to see if there is any issue :)

[psychofaktory]

In the meantime, I updated the Postgres driver and have been keeping a closer eye on the CPU usage.
After a very long initial scan (more than 24 hours) following the container's startup, the CPU usage is now actually only 0.01%.
A region update then takes place every 5 minutes, each lasting a good minute, even when no players are online.
During this time, the load is back up to about 16%.

Is it possible to extend the interval for the automatic region updates to a longer period (e.g., 1 hour or 24 hours)?
Alternatively, can the automatic updates be disabled entirely so that updates only occur when the container starts?

[TBlueF]

This sounds like you have a huge world with a lot of loaded chunks, or a slow database connection. Then this kind of behavior would be normal. Lots of changes or a slow database causing updates to take longer, bluemap has a harder time to keep up with the changing world.

Is it possible to extend the interval for the automatic region updates to a longer period (e.g., 1 hour or 24 hours)?

You are using the -u option when launching bluemap-cli, that means bluemap starts file-watchers on your world-files and each time the file-system detects that a region-file has changed, bluemap will start updating that region.
Updating means it will look at the timestamps of each chunk and compare it with the timestamp that the chunk had when bluemap last rendered it. -> If the timestamp changed, and only then, it actually renders the chunk.
That way bluemap only does work if there actually were changes.

The fact that the updates happens every 5 minutes for you, suggests that probably your minecraft server has a 5-minutes world-save interval and some loaded chunks. Even without players, maybe you have some chunk-loaders somewhere or some other mod/plugin that is loading chunks. Then whenever those chunks get saved, bluemap is forced to update them.

What you can do, there is a hidden setting in the plugin.conf, you can add update-cooldown: 3600 which would tell bluemap to update the same region-file never more than once an hour (3600 seconds).
So e.g. if a region is updated and then a change is detected in that same region-file after like 15 minutes then bluemap will schedule that region update to wait for another 45 minutes.
Note that this doesn't mean that there cant be more updates within that time, the cooldown is per region, so if different regions get changes, they will be updated immediately and then start their own cooldown.

Alternatively, can the automatic updates be disabled entirely so that updates only occur when the container starts?

Yes, just remove the -u flag from the container startup command (and use -r instead) :)

[psychofaktory]

Thank you very much for the detailed explanation.

My database connection shouldn't be the problem.
However, I do have a plugin active that keeps individual chunks loaded.
The advice regarding the update-cooldown seems to be exactly what I'm looking for. However, there is no plugin.conf file in the Docker version.
I have created the file manually. However, it doesn't seem to be taking effect. Is there another way to set up such a cooldown?

As a workaround, I’ve now removed the -u flag and could control the update intervals by having the container restart cyclically via a cron job.

[TBlueF]

You are right with that one, the plugin.conf is not available on cli...
I just pushed f7ec27d which moves the full-update-interval and the update-cooldown configs from the plugin.conf to the core.conf, i think they are suited there better anyways. So with the next release you will be able to use both of these also in the cli/docker version.

[Chicken]

Alternatively you can use them now by using the pre-release docker ghcr.io/bluemap-minecraft/bluemap:master