Basic commands

init

Create and initialize a ploop image file and a corresponding DiskDescriptor.xml file.

-s size

Image size. If no suffix is specified, the size is in sector units (one sector is 512 bytes). One can specify optional K, M, G or T suffix to set the size in kilo-, mega-, giga- or terabytes.

-f format

Image format. See Image formats below.

-v version

Image version, can be 1 or 2. Default is 2, if supported by the kernel.

-t none|ext3|ext4

File system type to create, default is ext4. Unless none is specified, a partition, a filesystem, and a balloon file will be created inside the image. Using ext3 is not recommended.

-b blocksize

Device block size, in 512 byte sectors. Default block size is 2048 sectors, or 1 megabyte.

-B fsblocksize

Filesystem block size, in bytes. Default is 4096 bytes.

-n, --nolazy

Disable lazy mkfs initialization (which is enabled by default). Currently that means if this flag is set, mkfs.ext4(8) is called with -Elazy_itable_init=0,lazy_journal_init=0 options.

delta_file

Path to a non-existent image file to be created.

mount

Assemble a ploop device from one or more delta images, start it, and optionally mount the file system residing on the device.

Two forms of this command are provided. The first one accepts a list of delta images to be used for assembling the ploop device, while the second one is using information from a DiskDescriptor.xml file. Please note that not all mount options are applicable to both forms.

-r

Mount as read-only.

-F

Run fsck(8) on inner filesystem before mounting it. This option is ignored if -m is not used.

-f format

Image format. Ignored if DiskDescriptor.xml is specified. Otherwise, one need to specify raw as an argument, if raw image format is used.

-b blocksize

Device block size, in 512 byte sectors. Ignored if DiskDescriptor.xml is specified. Otherwise, required for raw images.

-d device

Ploop device to use, e.g. /dev/ploop0. If not specified, a randomly numbered ploop device will be used.

-m mount_point

If this option is specified, ploop goes on to mount the file system to directory denoted by mount_point.

-o mount_options

Any additional mount options, comma-separated. Used if -m is set.

-t fstype

File system type used for mounting. Used if -m is set. The default is ext4.

-u uuid | base

GUID of the image from the DiskDescriptor.xml to be mounted. By default, top GUID is used. The special 'base' value can be used to mount the base (lower-level) image.

base_delta [... top_delta]

List of image files to mount, with the first one being the base delta and the last one being the top delta (i.e. the one that will be writable unless -r is specified).

DiskDescriptor.xml

Path to the DiskDescriptor.xml file with information about images.

umount

Unmount a ploop device. Since a mounted ploop device consists of an image (or multiple images), a device, and (optionally) a file system mounted to a directory, one can refer to any of the above entities to specify what to unmount. The recommended way is to use DiskDescriptor.xml.

-d device

Ploop device, e.g., /dev/ploop0.

-m mount_point

Mount point of a ploop device to unmount.

DiskDescriptor.xml

Path to the DiskDescriptor.xml file with information about images.

image_file

Path to a mounted image file.

replace

Replaces a ploop image by a different (but identical) one, on a running ploop device. Only a read-only image (e.g. a non-top one in a stacked configuration) can be replaced. An image to be replaced is specified by either one of level, UUID, or the current image file.

If a new image is not identical to the old one (i.e. its content differs) or not suitable for ploop in any other way (e.g. it is sparse, or resides on a file system not supported by ploop), the result is undefined.

-u uuid

A uuid of an image to be replaced.

-l level

A level of image to be replaced. A level is a distance of an image from the base delta.

-o cur_image_file

A current image file (the one to replace).

-k, --keep-name

A flag to keep the same image file name. If this flag is set, after the image is replaced, a new image_file is renamed to the old one (removing the old, now unused, image), so no modification to DiskDescriptor.xml is required.

-i image_file

A new replacement image.

DiskDescriptor.xml

Path to the DiskDescriptor.xml file with information about images.

resize

Resize a ploop image. Both online (i.e. when ploop is mounted and used) and offline resize is supported, and the tool can either grow or shrink both the ploop image and the underlying file system.

-s size

Image size. If no suffix is specified, size is in sector units (one sector is 512 bytes). One can specify optional K, M, G or T suffix to set the size in kilo-, mega-, giga- or terabytes.

DiskDescriptor.xml

Path to the DiskDescriptor.xml file with information about images.

convert

Convert either ploop image format or version (but not both at the same time). Conversion can only be performed offline (i.e. image should not be in use).

-f format

Image format. See Image formats below.

-v version

Image version, can be 1 or 2.

check

Check the internal consistency of (and possibly repair) a ploop image (or images). Note that image(s) to be tested should not be in use.

Check all the images in DiskDescriptor.xml up to the one denoted by the uuid (or default top delta, if UUID is not specified). Default built-in check options are used, and the ones specified on the command line, if any, are ignored.

-f, --force

Force check even if image's dirty flag is not set.

-F, --hard-force

Same as -f, plus try to fix even fatal errors (can be dangerous).

-c, --check

Check for duplicated blocks and holes.

-r, --ro

Read-only access, do not modify image(s).

-s, --silent

Be more silent, only report errors.

-d, --drop-inuse

Drop image "in use" flag.

-R, --raw

Specifies that image_file is a raw ploop image.

-b, --blocksize size

Image cluster block size, in sectors (for raw images).

-S, --repair-sparse

Repair sparse image(s).

Miscellaneous commands

info

list

Shows a list of running ploop devices (first column) and their corresponding base images. With option -a it also shows a mount point (third column).

restore-descriptor

Create DiskDescriptor.xml file suitable for delta_file and put it into disk_dir.

Read image header in case of ploop1 format or check raw image size and generate proper DiskDescriptor.xml file. You can specify blocksize for raw images. If it's not specified it will be chosen automatically - largest possible value between 32K and 1M. Raw image size must be aligned to blocksize.

This command works only for base images. Snapshots are not supported.

Working with snapshots

Ploop snapshots is a mechanism for creating and managing instant states of a running file system. Creating a snapshot leads to creating a new empty ploop image which is layered on top of an old one, then all writes are ending up in the top image, and reads are falling through to a lower level. There can be up to 126) stacked ploop images (or snapshots). Online snapshot merging is also supported.

Snapshots are identified by a unique UUID. A snapshot can be mounted using ploop mount -u uuid command, see above.

snapshot

Create a ploop snapshot.

-u uuid

Specify a uuid for a new snapshot. If option is not given, uuid is generated automatically. To generate uuid manually, one can use the uuidgen(1) utility. Note that UUID must be enclosed in curly brackets.

snapshot-merge

Merge a snapshot with its parent. That is, contents of the delta file corresponding to the snapshot is merged to a parent delta, then the file is removed. Parent snapshot UUID is lost (as it is replaced with the uuid specified). All snapshots having the lost one as a parent are updated to have the uuid as its parent.

-u uuid

Specify a single snapshot uuid to merge. If this option is not specified, the top delta will be used.

-U uuid2

Together with -u uuid, specify that all snapshots in the range uuid...uuid2 are to be merged.

-A

Merge all snapshots down to base delta. If some snapshots have more than a single child, they will be impossible to merge.

-n new_delta

If this option is set, instead of merging the child delta into its parent, both the parent and the child deltas are merged into a newly created file new_delta, which replaces the parent delta. Both deltas are then removed.

snapshot-switch

Switch to the specified snapshot. This operation can only be performed while ploop is not running (i.e. is unmounted). The current top delta image will be removed.

-u uuid

Specify a snapshot uuid to switch to.

snapshot-delete

Delete the specified snapshot. This operation can only be performed if the specified snapshot is not active. In case snapshot doesn't have any children, it will simply be removed. In case snapshot has a single child, it will be merged to that child. Deleting a snapshot that has multiple children is currently not supported (but can be performed manually in an iterative fashion).

-u uuid

Specify a snapshot uuid to be deleted.

snapshot-list

List available snapshots.

-H, --no-header

Suppress displaying the header row. Usable for scripts.

-u, --uuid, --id uuid

Filter the output to a specified uuid.

-s, --snapshot

List in terms of snapshots. By default (i.e. without this option) the command lists all deltas, and top delta is marked as current. When this option is used, top delta is not listed, and the snapshot which is the parent of top delta is marked as current.

-o, --output field[,field...]

Display only the specified fields. Possible fields are:
• uuid - snapshot's UUID;
• parent_uuid - snapshot's parent UUID;
• current - if this snapshot is the current one;
• fname - snapshot image file name.

Image copying

ploop copy is a mechanism of effective copying of a top ploop image with the help of build-in ploop kernel driver feature called write tracker. Write tracker is a feature that lets ploop copy to iteratively obtain a list of modified image blocks from the kernel. Two ploop copy processes are required for iterative top delta transfer. These are used by vzmigrate(8).

copy (sending)

This command enables the in-kernel write tracker for the specified ploop device, then sends all the data blocks from the top delta image to a pipe specified by the output_fd argument (stdout, i.e. 1 by default), supposedly read by destination ploop copy, or a file. After that, it iteratively gets the list of the modified data blocks from the kernel and sends those blocks again. After a number of iterations (or when the list is empty), it executes the stop_command (this could be vzctl stop or vzctl chkpnt) and does the last iteration of sending the modified data blocks. Finally, it checks that the data were not modified, error is returned otherwise.

If feedback_fd is specified, it is used to read back from the ploop copy receiving side. The feedback channel is currently used to wait for fdatasync(2) completion.

copy (receiving)

Reads the data blocks (provided by the source ploop copy) from the file descriptor input_fd (stdin, i.e. 0 by default) and writes them to the file.

If feedback_fd is specified, it is used to send status back to the ploop copy sending side.

Ballooning

Since there is no online shrink support in ext4 file system, ploop internally uses a technique called "ballooning" as a work around to shrink its images.

Ballooning operation consists of inflating a special balloon file (invisible for ordinary users), loading fiemap info of the inflated balloon to the kernel, relocating blocks of the image file from the tail to the space specified by fiemap info, and truncating the tail of the image file. Result is the image file of a smaller size.

However, it is quite possible that inflated balloon file will only span blocks that were never touched before. Those will look like "not allocated" space from the kernel ploop point of view. In this case nothing will be relocated and nothing truncated.

So, if balloon operation succeeded, it's only guaranteed that a user of ploop device won't be able to consume more space than the initial block device size minus the size of the inflated balloon. On the other hand, if a user of block device used a lot of space on it, then freed the significant part of used space, balloon operation will result in significant truncate of image file.

All the ploop ballooning logic is hidden from the end user, so while a number of low-level commands exist for working with ploop ballooning, those are not needed and therefore are not documented here, except for a single command.

balloon discard

In a situation when a lot of disk space were freed on an in-ploop filesystem, use ploop balloon discard to optimize the ploop image size.

Iteratively try to relocate and discard unused blocks from a ploop image, reducing its size.

Note that ploop device and its inner file system should be mounted. If not, one can use --automount option to automatically mount ploop for the duration of the operation.

Option --defrag can be used to run a filesystem defragmentation utility (currently e4defrag2 on ext4 only) before the main operation.

Option --to-free can be used to specify a maximum disk space to be freed. In other words, stop the process once freed space exceeded requested size. Default is 0, meaning to try to free as much space as possible.

Option --min-block can be used to specify a minimum size of an extent to free. The smallest possible extent is 1 cluster (currently 1 MB), one can specify higher value to speed up the whole discarding operation.

Note that the same functionality is available by means of vzctl compact command.

Image formats

The following image formats are currently supported.

raw

Raw format, with 1:1 mapping between the image file and the ploop device.

ploop1, expanded

Expanded format. The image will grow according to the needs of the underlying file system. This format is the default. Names 'ploop1' and 'expanded' are aliases.

preallocated

This is the same as 'ploop1' or 'expanded', the only difference is all the file blocks are allocated during creation.