Default limit for core dumps of balena engine (and thus all its containers) is too high

[jellyfish-bot]

[hades32] Several customers had full disks because of too many core dumps. There is no reason to allow more than a few MB. (the number is in 1K units)

root@86fa459:~# grep Limit /lib/systemd/system/balena.service 
LimitNOFILE=1048576
LimitNPROC=1048576
LimitCORE=infinity # this

related: moby/moby#38814

[Hades32]

@robertgzr any reason not to change this limit?

[robertgzr]

@Hades32 there's an issue on meta-balena that we should review the systemd service: balena-os/meta-balena#1910

although my plan was to bring the upstream service file (https://github.com/moby/moby/blob/master/contrib/init/systemd/docker.service) into balenaOS, which would make all these limits infinity

[Hades32]

@robertgzr but why? Shouldn't we explicitly set it to a low number and NOT to infinity?

[robertgzr]

I can't think of anything against lowering them... maybe that the SV would also be affected, since they apply to all containers. cc @alexgg wdyt regaring ulimits on containers?

[Hades32]

@robbyt I realize it was confusing that I added the two other limits in the code snippet above. I was specifically talking about LimitCORE. I don't think we have regular use for core dumps...

[alexgg]

@robertgzr I think this is something we could add to a developmentMode image and remove in a production image.

[jellyfish-bot]

[kb2ma] This issue has attached support thread https://jel.ly.fish/f50e9cf8-d51b-4000-b5ed-84f8cae452cc

[Petersdavis]

I found that this is the issue causing our devices to brick themselves.