Understanding GrootFS Disk Usage

Page last updated:

This topic explains the concepts related to GrootFS disk space management in Pivotal Web Services (PWS).

GrootFS Stores

GrootFS is the container root filesystem management component for Garden. A container root filesystem or rootfs is often referred to as an image.

A GrootFS store is the directory in which rootfs layers and container images are cached. This directory is configured by GrootFS and mounted on an XFS-formatted volume by the Garden job during BOSH VM creation.

Individual container root filesystems are provided via OverlayFS mounts.

Supplying GrootFS with an already formatted XFS volume for its store is not yet supported for BOSH-controlled deployments.

General Garbage Collection Behavior in GrootFS Stores

GrootFS stores are initialized to use the entirety of /var/vcap/data. If the reserved_space_for_other_jobs_in_mb is not set high enough, or if there are many images with few shared volumes, the store can use up all available space.

The thresholder component calculates and sets a value so that GrootFS’s garbage collector can attempt to ensure that a small reserved space is kept free for other jobs. GrootFS only tries to garbage collect when that threshold is reached. However, if all the rootfs layers are actively in use by images, then garbage collection cannot occur and that space is used up.

Volumes

Underlying layers in rootfs images are known as volumes in GrootFS. They are read-only and their changesets are layered together through an OverlayFS mount to create the root filesystems for containers.

When GrootFS writes each filesystem volume to disk, it also stores the number of bytes written to a file in a meta directory. The size of an individual volume is available in its corresponding metadata file. GrootFS also stores the SHA of each underlying volume used by an image in the meta folder.

For each container, GrootFS mounts the underlying volumes using overlay to a point in the images directory. This mount point is the rootfs for the container and is read-write.

On disk, the read-write layer for each container can be found at /var/vcap/data/grootfs/store/unprivileged/images/CONTAINER-ID/diff (or /var/vcap/data/grootfs/store/privileged/images/CONTAINER-ID/diff for privileged containers.)

When GrootFS calls on the built-in XFS quota tooling to get disk usage for a container, it takes into account data written to those diff directories and not the data in the read-only volumes.

Volume Cleanup Example

When clean is called in GrootFS, any layers that are not being used by an existing rootfs are deleted from the store. The cleanup only takes into account the volumes folders in the store.

For example, imagine that there are two rootfs images from different base images, Image A and Image B:

- Image A
  Layers:
    - layer-1
    - layer-2
    - layer-3

- Image B
  Layers:
    - layer-1
    - layer-4
    - layer-5

They have a layer in common, layer-1. And after deleting Image B, layer-4 and layer-5 can be collected by clean, but not layer-1 because Image A still uses that layer.

Additional Information

For more information, see the following sections of garden-runc-release:

Create a pull request or raise an issue on the source for this page in GitHub