garden-linux/0.334.0

matching garden sha: 518f19c4f33b5ac1314c2c5c52c1190dc1508b20

Breaking property change: The enable_graph_cleanup property is removed, and is replaced by the new graph_cleanup_threshold_in_mb property. If this new property is not set, graph cleanup will be disabled.

Previously, if graph cleanup was enabled, unused filesystem layers would be deleted every time a container was created. It’s now possible to specify a threshold for the amount of disk space that needs to be used up by filesystem layers downloaded from Docker registries before graph cleanup will run. This allows deployments to be configured such that unused filesystem layers will only be deleted when the system is potentially under disk pressure. The advantage of this is that running Docker images that were recently running, but have since been deleted, will be much faster since there’s no need to re-download those images from the registry.

To take advantage of this, set graph_cleanup_threshold_in_mb to a positive integer - graph cleanup will be triggered on container creation when the total size of all downloaded Docker image layers exceeds this size. If you’d like to retain the old behaviour, pick the option below that applies to you: - If you previously didn’t specify enable_graph_cleanup, you don’t need to do anything - the default behaviour has not changed - If you previously set enable_graph_cleanup to true, and want to retain this behaviour, set graph_cleanup_threshold_in_mb to 0 - If you previously set enable_graph_cleanup to false, and want to retain this behaviour, set graph_cleanup_threshold_in_mb to -1, or simply don’t specify either - the default is -1

Also in this release: - Fix a bug where creating or deleting containers could cause other containers to temporarily lose network connectivity