armhf alpine - docker container high CPU usage about docker-mariadb

[lippl]

For Reference: I am having the same issue and found the following Bug Reports already filed at Alpine and MariaDB

[lippl]

Well, I just edited the dockerfile to built using FROM ghcr.io/linuxserver/baseimage-alpine:arm32v7-3.12 and the argument MARIADB_VERSION: 10.4.17-r1. That is working for now.
The bugs also mention the issue to be fixed on the latest edge version of alpine. But for now I am just waiting for an updated version here. Until then I will stick with the older 3.12/10.4 version.

[aptalca]

I can reproduce on arm32. Doesn't affect aarch64. Could be (likely is) an upstream issue. Will investigate further.

[nemchik]

Can anyone who previously confirmed this issue have a look at #81 and try the tag in the OP to see if they have issues with it?

[nishantbb]

Same thing on my end.

Same here. Added some of the info requested on the MariaDB bug tracker to hopefully keep things pushing along.

I saw on that tracker that the issue is not occurring for 10.4.x, are these versions possible with the alpine image?

[lippl]

Thanks, Looking good so far from my side!

[ajaskiewiczpl]

I know this might not be a solution for everyone, but what I can recommend for Raspberry Pi users is to install 64-bit Ubuntu Server. There are lot more official docker images for ARM64 platform. I am running it since a week and everything works perfectly fine.

[Bolkarr]

Unfortunately not everyone has a 64bit capable system. For example I use a Odroid HC2. The whole system is under constant high CPU use, which is preventing it to use resources for other applications, we desperately need a fix for this...

[ajaskiewiczpl]

I understand, that's why I said this is not a solution for everyone. Another "workaround" that might work for you, is to temporarily limit CPU resources for docker container until the fix is ready.
--cpus="0.25"
Again, this is not a solution but rather temporary workaround, that at least give more cpu resources for other apps in the system.

[w4z]

I have the same issue. Limiting temporarily CPU via --cpus= does not help too much as everything becomes too slow.

[iDaoU]

Well, my solution is to use the tobi312/rpi-mariadb:10.4-alpine docker image. From here.

I think the long-term solution for linuxserver/docker-mariadb would be:

1. Upgrade the base image to Ubuntu Focal (mariadb:10.3), or
2. Wait for Alpine 3.13's mariadb:10.5+ setup to become stable, or
3. (Edit after suggestion from @Bolkarr) Use Alpine 3.12 and mariadb:10.4 for the Alpine branch/image.

I personally think the first (or third) option would have a more controllable timeline.

[iDaoU]

@Bolkarr
Sorry I misunderstood your question earlier. I don't know if the volume mappings are compatible. tobi312's mappings are as below:

volumes:
  - /mariadb:/var/lib/mysql

[aptalca]

Please refrain from recommending random images that are not compatible. That will only confuse users further. Let's keep this issue on topic.

Thanks.

[iDaoU]

Tag is lspipepr/mariadb:arm32v7-alpine-10.5.9-r0-pkg-6271d700-pr-81 for lazy people.

[joshtrichards]

Same here. Added some of the info requested on the MariaDB bug tracker to hopefully keep things pushing along.

[ridgarou]

Here another one with the same problem. :(

[Bolkarr]

Yes, same issue with the alpine on armhf.

[sunrisepi]

I know this might not be a solution for everyone, but what I can recommend for Raspberry Pi users is to install 64-bit Ubuntu Server. There are lot more official docker images for ARM64 platform. I am running it since a week and everything works perfectly fine.

You are so right! I have had to make my own Docker images for some apps (Wallabag is one example) as there is no 32-bit ARM version. I have been considering moving from 32- to 64-bit OS, but it's a lot of work.

[Bolkarr]

Is tobi312 interchangeable with the alpine version? (can we just recreate the docker with this image?)

[Bolkarr]

Yes, it seems like the high CPU usage is gone. CPU wise it is working as it should. :)

[jwillmer]

I can confirm as well that the referenced image is working. On a Pi4 the load went done to .04%

[aptalca]

#81 merged