[question] graph_cleanup_threshold_in_mb about garden-runc-release HOT 7 CLOSED

Ok, thanks a lot @glestaris !

from garden-runc-release.

We have created an issue in Pivotal Tracker to manage this:

https://www.pivotaltracker.com/story/show/150631625

The labels on this github issue will be updated when the story is started.

from garden-runc-release.

Is the cleanup process activated once the /var/vcap/data/garden/graph/aufs/diff/ folder reaches 65 GB? If yes, then it's normal the cleanup process didn't run. It's 64GB now.

from garden-runc-release.

Hello @paolostivanin. Yes, unfortunately, the threshold only takes into account the sum of the sizes of the layers. 65 GB is a lot for an 83 GB disk so cleanup never runs.

Depending on the workload your cells are running, a lower threshold (such as 1000-2000 MB) will occasionally trigger the cleanup process.

from garden-runc-release.

GrootFS is doing some work to remove the threshold property and automatically trigger cleanup based on remaining disk space.

from garden-runc-release.

Ok, thanks! But wouldn't a 2G value be too low? This would cause the cleanup process to run too often, right? Is this ok or can it cause some issues?

from garden-runc-release.

It really depends on your workload. 1-2G would be a good threshold if you are mainly using buildpack-based apps (as opposed to Docker-based apps).

The GC process is fairly lightweight so even if it was very frequent it wouldn't significantly affect other cell operations. That's mainly because if run frequently it won't find layers to delete all the time so all it does is in these cases is check metadata.

from garden-runc-release.