QEMU Block Layer currently (as of QEMU 2.9) supports four major kinds of live block device jobs -- stream, commit, mirror, and backup. These can be used to manipulate disk image chains to accomplish certain tasks, namely: live copy data from backing files into overlays; shorten long disk image chains by merging data from overlays into backing files; live synchronize data from a disk image chain (including current active disk) to another target image; point-in-time (and incremental) backups of a block device. Below is a description of the said block (QMP) primitives, and some (non-exhaustive list of) examples to illustrate their use.
Note
The file qapi/block-core.json in the QEMU source tree has the canonical QEMU API (QAPI) schema documentation for the QMP primitives discussed here.
Contents
A simple disk image chain. (This can be created live, using QMP blockdev-snapshot-sync, or offline, via qemu-img):
(Live QEMU)
|
.
V
[A] <----- [B]
(backing file) (overlay)
The arrow can be read as: Image [A] is the backing file of disk image [B]. And live QEMU is currently writing to image [B], consequently, it is also referred to as the "active layer".
There are two kinds of terminology that are common when referring to files in a disk image backing chain:
Directional: 'base' and 'top'. Given the simple disk image chain above, image [A] can be referred to as 'base', and image [B] as 'top'. (This terminology can be seen in in QAPI schema file, block-core.json.)
Relational: 'backing file' and 'overlay'. Again, taking the same simple disk image chain from the above, disk image [A] is referred to as the backing file, and image [B] as overlay.
Throughout this document, we will use the relational terminology.
Important
The base disk image can be raw format; however, all the overlay files must be of QCOW2 format.
The following are the four different kinds of live block operations that QEMU block layer supports.
To show some example invocations of command-line, we will use the following invocation of QEMU, with a QMP server running over UNIX socket:
$ ./x86_64-softmmu/qemu-system-x86_64 -display none -nodefconfig \
-M q35 -nodefaults -m 512 \
-blockdev node-name=node-A,driver=qcow2,file.driver=file,file.node-name=file,file.filename=./a.qcow2 \
-device virtio-blk,drive=node-A,id=virtio0 \
-monitor stdio -qmp unix:/tmp/qmp-sock,server,nowait
The -blockdev command-line option, used above, is available from QEMU 2.9 onwards. In the above invocation, notice the 'node-name' parameter that is used to refer to the disk image a.qcow2 ('node-A') -- this is a cleaner way to refer to a disk image (as opposed to referring to it by spelling out file paths). So, we will continue to designate a 'node-name' to each further disk image created (either via blockdev-snapshot-sync, or blockdev-add) as part of the disk image chain, and continue to refer to the disks using their 'node-name' (where possible, because block-stream, and block-commit do not yet, as of QEMU 2.9, take 'node-name' parameters) when performing various block operations.
To interact with the QEMU instance launched above, we will use the qmp-shell (located at: qemu/scripts/qmp, as part of the QEMU source directory) utility, which takes key-value pairs for QMP commands. Invoke it as below (which will also print out the complete raw JSON syntax for reference -- examples in the following sections):
$ ./qmp-shell -v -p /tmp/qmp-sock (QEMU)
Note
In the event we have to repeat a certain QMP command, we will: for the first occurrence of it, show the the qmp-shell invocation, and the corresponding raw JSON QMP syntax; but for subsequent invocations, present just the qmp-shell syntax, and omit the equivalent JSON output.
We will use the below disk image chain (and occasionally spelling it out where appropriate) when discussing various primitives:
[A] <-- [B] <-- [C] <-- [D]
Where [A] is the original base image; [B] and [C] are intermediate overlay images; image [D] is the active layer -- i.e. live QEMU is writing to it. (The rule of thumb is: live QEMU will always be pointing to the right-most image in a disk image chain.)
The above image chain can be created by invoking blockdev-snapshot-sync command as following (which shows the creation of overlay image [B]) using the qmp-shell (our invocation also prints the raw JSON invocation of it):
(QEMU) blockdev-snapshot-sync node-name=node-A snapshot-file=b.qcow2 snapshot-node-name=node-B format=qcow2
{
"execute": "blockdev-snapshot-sync",
"arguments": {
"node-name": "node-A",
"snapshot-file": "b.qcow2",
"format": "qcow2",
"snapshot-node-name": "node-B"
}
}
Here, "node-A" is the name QEMU internally uses to refer to the base image [A] -- it is the backing file, based on which the overlay image, [B], is created.
To create the rest of the two overlay images, [C], and [D] (omitted the raw JSON output for brevity):
(QEMU) blockdev-snapshot-sync node-name=node-B snapshot-file=c.qcow2 snapshot-node-name=node-C format=qcow2 (QEMU) blockdev-snapshot-sync node-name=node-C snapshot-file=d.qcow2 snapshot-node-name=node-D format=qcow2
In our disk image chain:
[A] <-- [B] <-- [C] <-- [D]
We have three points in time and an active layer:
Therefore, be aware with naming choices:
The block-stream command allows you to do live copy data from backing files into overlay images.
Given our original example disk image chain from earlier:
[A] <-- [B] <-- [C] <-- [D]
The disk image chain can be shortened in one of the following different ways (not an exhaustive list).
Merge everything into the active layer: I.e. copy all contents from the base image, [A], and overlay images, [B] and [C], into [D], _while_ the guest is running. The resulting chain will be a standalone image, [D] -- with contents from [A], [B] and [C] merged into it (where live QEMU writes go to):
[D]
Taking the same example disk image chain mentioned earlier, merge only images [B] and [C] into [D], the active layer. The result will be contents of images [B] and [C] will be copied into [D], and the backing file pointer of image [D] will be adjusted to point to image [A]. The resulting chain will be:
[A] <-- [D]
Intermediate streaming (available since QEMU 2.8): Starting afresh with the original example disk image chain, with a total of four images, it is possible to copy contents from image [B] into image [C]. Once the copy is finished, image [B] can now be (optionally) discarded; and the backing file pointer of image [C] will be adjusted to point to [A]. I.e. after performing "intermediate streaming" of [B] into [C], the resulting image chain will be (where live QEMU is writing to [D]):
[A] <-- [C] <-- [D]
For Case-1, to merge contents of all the backing files into the active layer, where 'node-D' is the current active image (by default block-stream will flatten the entire chain); qmp-shell (and its corresponding JSON output):
(QEMU) block-stream device=node-D job-id=job0
{
"execute": "block-stream",
"arguments": {
"device": "node-D",
"job-id": "job0"
}
}
For Case-2, merge contents of the images [B] and [C] into [D], where image [D] ends up referring to image [A] as its backing file:
(QEMU) block-stream device=node-D base-node=node-A job-id=job0
And for Case-3, of "intermediate" streaming", merge contents of images [B] into [C], where [C] ends up referring to [A] as its backing image:
(QEMU) block-stream device=node-C base-node=node-A job-id=job0
Progress of a block-stream operation can be monitored via the QMP command:
(QEMU) query-block-jobs
{
"execute": "query-block-jobs",
"arguments": {}
}
Once the block-stream operation has completed, QEMU will emit an event, BLOCK_JOB_COMPLETED. The intermediate overlays remain valid, and can now be (optionally) discarded, or retained to create further overlays based on them. Finally, the block-stream jobs can be restarted at anytime.
The block-commit command lets you to live merge data from overlay images into backing file(s). Since QEMU 2.0, this includes "live active commit" (i.e. it is possible to merge the "active layer", the right-most image in a disk image chain where live QEMU will be writing to, into the base image). This is analogous to block-stream, but in opposite direction.
Again, starting afresh with our example disk image chain, where live QEMU is writing to the right-most image in the chain, [D]:
[A] <-- [B] <-- [C] <-- [D]
The disk image chain can be shortened in one of the following ways:
Commit content from only image [B] into image [A]. The resulting chain is the following, where image [C] is adjusted to point at [A] as its new backing file:
[A] <-- [C] <-- [D]
Commit content from images [B] and [C] into image [A]. The resulting chain, where image [D] is adjusted to point to image [A] as its new backing file:
[A] <-- [D]
Commit content from images [B], [C], and the active layer [D] into image [A]. The resulting chain (in this case, a consolidated single image):
[A]
Commit content from image only image [C] into image [B]. The resulting chain:
[A] <-- [B] <-- [D]
Commit content from image [C] and the active layer [D] into image [B]. The resulting chain:
[A] <-- [B]
For block-commit_Case-1, to merge contents only from image [B] into image [A], the invocation is as following:
(QEMU) block-commit device=node-D base=a.qcow2 top=b.qcow2 job-id=job0
{
"execute": "block-commit",
"arguments": {
"device": "node-D",
"job-id": "job0",
"top": "b.qcow2",
"base": "a.qcow2"
}
}
Once the above block-commit operation has completed, a BLOCK_JOB_COMPLETED event will be issued, and no further action is required. The end result being, the backing file of image [C] is adjusted to point to image [A], and the original 4-image chain will end up being transformed to:
[A] <-- [C] <-- [D]
However, block-commit_Case-3 (also called: "active block-commit") is a two-phase operation: in the first phase, the content from the active overlay, along with the intermediate overlays, is copied into the backing file (also called, the base image); in the second phase, adjust the said backing file as the current active image -- possible via issuing the command block-job-complete. [Optionally, the block-commit operation can be cancelled, by issuing the command block-job-cancel, but be careful when doing this.]
Once the 'commit' operation (started by block-commit) has completed, the event BLOCK_JOB_READY is emitted, signalling the synchronization has finished, and the job can be gracefully completed, by issuing block-job-complete. (Until such a command is issued, the 'commit' operation remains active.)
So, the following is the flow for block-commit_Case-3 -- to convert a disk image chain such as this:
[A] <-- [B] <-- [C] <-- [D]
Into (where content from all the subsequent overlays, [B], and [C], including the active layer, [D], is committed back to [A] -- which is where live QEMU is performing all its current writes):
[A]
Start the "active block-commit" operation:
(QEMU) block-commit device=node-D base=a.qcow2 top=d.qcow2 job-id=job0
{
"execute": "block-commit",
"arguments": {
"device": "node-D",
"job-id": "job0",
"top": "d.qcow2",
"base": "a.qcow2"
}
}
Once the synchronization has completed, the event BLOCK_JOB_READY will be emitted.
Then, (optionally) query for the status of the active block operations (we can see the 'commit' job is now ready to be completed, as indicated by the line "ready": true):
(QEMU) query-block-jobs
{
"execute": "query-block-jobs",
"arguments": {}
}
{
"return": [
{
"busy": false,
"type": "commit",
"len": 1376256,
"paused": false,
"ready": true,
"io-status": "ok",
"offset": 1376256,
"device": "job0",
"speed": 0
}
]
}
Gracefully, complete the 'commit' block device job:
(QEMU) block-job-complete device=job0
{
"execute": "block-job-complete",
"arguments": {
"device": "job0"
}
}
{
"return": {}
}
Finally, once the above job is completed, an event BLOCK_JOB_COMPLETED will be emitted.
Note
The invocation for rest of the cases (2, 4, and 5), discussed in the previous section, is omitted for brevity.
Synchronize a running disk image chain (all or part of it) to a target image.
Again, given our familiar disk image chain:
[A] <-- [B] <-- [C] <-- [D]
The drive-mirror (and its newer equivalent blockdev-mirror) allows you to copy data from the entire chain into a single target image (which can be located on a different host).
Once a 'mirror' job has started, there are two possible actions when a drive-mirror job is active:
To copy the contents of the entire disk image chain, from [A] all the way to [D], to a new target (drive-mirror will create the destination file, if it doesn't already exist), call it [E]:
(QEMU) drive-mirror device=node-D target=e.qcow2 sync=full job-id=job0
{
"execute": "drive-mirror",
"arguments": {
"device": "node-D",
"job-id": "job0",
"target": "e.qcow2",
"sync": "full"
}
}
The "sync": "full", from the above, means: copy the entire chain to the destination.
Following the above, querying for active block jobs will show that a 'mirror' job is "ready" to be completed (and QEMU will also emit an event, BLOCK_JOB_READY):
(QEMU) query-block-jobs
{
"execute": "query-block-jobs",
"arguments": {}
}
{
"return": [
{
"busy": false,
"type": "mirror",
"len": 21757952,
"paused": false,
"ready": true,
"io-status": "ok",
"offset": 21757952,
"device": "job0",
"speed": 0
}
]
}
And, as mentioned in the previous section, the two possible options can be taken:
Create a point-in-time snapshot by ending the synchronization. The point-in-time is at the time of ending the sync. (The result of the following being: the target image, [E], will be populated with content from the entire chain, [A] to [D]):
(QEMU) block-job-cancel device=job0
{
"execute": "block-job-cancel",
"arguments": {
"device": "job0"
}
}
Or, complete the operation and pivot the live QEMU to the target copy:
(QEMU) block-job-complete device=job0
In either of the above cases, if you once again run the query-block-jobs command, there should not be any active block operation.
Comparing 'commit' and 'mirror': In both then cases, the overlay images can be discarded. However, with 'commit', the existing base image will be modified (by updating it with contents from overlays); while in the case of 'mirror', a new target image is populated with the data from the disk image chain.
Live storage migration (without shared storage setup) is one of the most common use-cases that takes advantage of the drive-mirror primitive and QEMU's built-in Network Block Device (NBD) server. Here's a quick walk-through of this setup.
Given the disk image chain:
[A] <-- [B] <-- [C] <-- [D]
Instead of copying content from the entire chain, synchronize only the contents of the top-most disk image (i.e. the active layer), [D], to a target, say, [TargetDisk].
Important
The destination host must already have the contents of the backing chain, involving images [A], [B], and [C], visible via other means -- whether by cp, rsync, or by some storage array-specific command.)
Sometimes, this is also referred to as "shallow copy" -- because: only the "active layer", and not the rest of the image chain, is copied to the destination.
Note
In this, for the sake of simplicity, we'll be using the same local host as both, source and destination.
As noted earlier, on the destination host the contents of the backing chain -- from images [A] to [C] -- are already expected to exist in some form (e.g. in a file called, Contents-of-A-B-C.qcow2). Now, on the destination host, let's create a target overlay image (with the image Contents-of-A-B-C.qcow2 as its backing file), to which the contents of image [D] (from the source QEMU) will be mirrored to:
$ qemu-img create -f qcow2 -b ./Contents-of-A-B-C.qcow2 \
-F qcow2 ./target-disk.qcow2
And start the destination QEMU (we already have the source QEMU running -- discussed in the section: Interacting with a QEMU instance) instance, with the following invocation. (As noted earlier, for simplicity's sake, the destination QEMU is started on the same host, but it could be located elsewhere):
$ ./x86_64-softmmu/qemu-system-x86_64 -display none -nodefconfig \
-M q35 -nodefaults -m 512 \
-blockdev node-name=node-TargetDisk,driver=qcow2,file.driver=file,file.node-name=file,file.filename=./target-disk.qcow2 \
-device virtio-blk,drive=node-TargetDisk,id=virtio0 \
-S -monitor stdio -qmp unix:./qmp-sock2,server,nowait \
-incoming tcp:localhost:6666
Given the disk image chain on source QEMU:
[A] <-- [B] <-- [C] <-- [D]
On the destination host, it is expected that the contents of the chain [A] <-- [B] <-- [C] are already present, and therefore copy only the content of image [D].
[On destination QEMU] As part of the first step, start the built-in NBD server on a given host (local host, represented by ::)and port:
(QEMU) nbd-server-start addr={"type":"inet","data":{"host":"::","port":"49153"}}
{
"execute": "nbd-server-start",
"arguments": {
"addr": {
"data": {
"host": "::",
"port": "49153"
},
"type": "inet"
}
}
}
[On destination QEMU] And export the destination disk image using QEMU's built-in NBD server:
(QEMU) nbd-server-add device=node-TargetDisk writable=true
{
"execute": "nbd-server-add",
"arguments": {
"device": "node-TargetDisk"
}
}
[On source QEMU] Then, invoke drive-mirror (NB: since we're running drive-mirror with mode=existing (meaning: synchronize to a pre-created file, therefore 'existing', file on the target host), with the synchronization mode as 'top' ("sync: "top"):
(QEMU) drive-mirror device=node-D target=nbd:localhost:49153:exportname=node-TargetDisk sync=top mode=existing job-id=job0
{
"execute": "drive-mirror",
"arguments": {
"device": "node-D",
"mode": "existing",
"job-id": "job0",
"target": "nbd:localhost:49153:exportname=node-TargetDisk",
"sync": "top"
}
}
[On source QEMU] Once drive-mirror copies the entire data, and the event BLOCK_JOB_READY is emitted, issue block-job-cancel to gracefully end the synchronization, from source QEMU:
(QEMU) block-job-cancel device=job0
{
"execute": "block-job-cancel",
"arguments": {
"device": "job0"
}
}
[On destination QEMU] Then, stop the NBD server:
(QEMU) nbd-server-stop
{
"execute": "nbd-server-stop",
"arguments": {}
}
[On destination QEMU] Finally, resume the guest vCPUs by issuing the QMP command cont:
(QEMU) cont
{
"execute": "cont",
"arguments": {}
}
Note
Higher-level libraries (e.g. libvirt) automate the entire above process.
The blockdev-mirror command is equivalent in core functionality to drive-mirror, except that it operates at node-level in a BDS graph.
Also: for blockdev-mirror, the 'target' image needs to be explicitly created (using qemu-img) and attach it to live QEMU via blockdev-add, which assigns a name to the to-be created target node.
E.g. the sequence of actions to create a point-in-time backup of an entire disk image chain, to a target, using blockdev-mirror would be:
Given the disk image chain:
[A] <-- [B] <-- [C] <-- [D]
To copy the contents of the entire disk image chain, from [A] all the way to [D], to a new target, call it [E]. The following is the flow.
Create the overlay images, [B], [C], and [D]:
(QEMU) blockdev-snapshot-sync node-name=node-A snapshot-file=b.qcow2 snapshot-node-name=node-B format=qcow2 (QEMU) blockdev-snapshot-sync node-name=node-B snapshot-file=c.qcow2 snapshot-node-name=node-C format=qcow2 (QEMU) blockdev-snapshot-sync node-name=node-C snapshot-file=d.qcow2 snapshot-node-name=node-D format=qcow2
Create the target image, [E]:
$ qemu-img create -f qcow2 e.qcow2 39M
Add the above created target image to QEMU, via blockdev-add:
(QEMU) blockdev-add driver=qcow2 node-name=node-E file={"driver":"file","filename":"e.qcow2"}
{
"execute": "blockdev-add",
"arguments": {
"node-name": "node-E",
"driver": "qcow2",
"file": {
"driver": "file",
"filename": "e.qcow2"
}
}
}
Perform blockdev-mirror, and observe for the event BLOCK_JOB_READY:
(QEMU) blockdev-mirror device=node-B target=node-E sync=full job-id=job0
{
"execute": "blockdev-mirror",
"arguments": {
"device": "node-D",
"job-id": "job0",
"target": "node-E",
"sync": "full"
}
}
Query for active block jobs, there should be a 'mirror' job ready:
(QEMU) query-block-jobs
{
"execute": "query-block-jobs",
"arguments": {}
}
{
"return": [
{
"busy": false,
"type": "mirror",
"len": 21561344,
"paused": false,
"ready": true,
"io-status": "ok",
"offset": 21561344,
"device": "job0",
"speed": 0
}
]
}
Gracefully complete the block device job operation, and observe for the event BLOCK_JOB_COMPLETED:
(QEMU) block-job-complete device=job0
{
"execute": "block-job-complete",
"arguments": {
"device": "job0"
}
}
{
"return": {}
}
Shutdown the guest, by issuing the quit QMP command:
(QEMU) quit
{
"execute": "quit",
"arguments": {}
}
The drive-backup (and its newer equivalent blockdev-backup) allows you to create a point-in-time snapshot.
In this case, the point-in-time is when you start the drive-backup (or its newer equivalent blockdev-backup) command.
Yet again, starting afresh with our example disk image chain:
[A] <-- [B] <-- [C] <-- [D]
To create a target image [E], with content populated from image [A] to [D], from the above chain, the following is the syntax. (If the target image does not exist, drive-backup will create it):
(QEMU) drive-backup device=node-D sync=full target=e.qcow2 job-id=job0
{
"execute": "drive-backup",
"arguments": {
"device": "node-D",
"job-id": "job0",
"sync": "full",
"target": "copy-drive-backup.qcow2"
}
}
Once the above drive-backup has completed, a BLOCK_JOB_COMPLETED event will be issued, indicating the live block device job operation has completed, and no further action is required.
The blockdev-backup command is equivalent in functionality to drive-backup, except that it operates at node-level in a Block Driver State (BDS) graph.
E.g. the sequence of actions to create a point-in-time backup of an entire disk image chain, to a target, using blockdev-backup would be:
The following section shows an example QMP invocation for blockdev-backup.
Given, a disk image chain of depth 1, where image [B] is the active overlay (live QEMU is writing to it):
[A] <-- [B]
The following is the procedure to copy the content from the entire chain to a target image (say, [E]), which has the full content from [A] and [B].
Create the overlay, [B]:
(QEMU) blockdev-snapshot-sync node-name=node-A snapshot-file=b.qcow2 snapshot-node-name=node-B format=qcow2
{
"execute": "blockdev-snapshot-sync",
"arguments": {
"node-name": "node-A",
"snapshot-file": "b.qcow2",
"format": "qcow2",
"snapshot-node-name": "node-B"
}
}
Create a target image, that will contain the copy:
$ qemu-img create -f qcow2 e.qcow2 39M
Then, add it to QEMU via blockdev-add:
(QEMU) blockdev-add driver=qcow2 node-name=node-E file={"driver":"file","filename":"e.qcow2"}
{
"execute": "blockdev-add",
"arguments": {
"node-name": "node-E",
"driver": "qcow2",
"file": {
"driver": "file",
"filename": "e.qcow2"
}
}
}
Then, invoke blockdev-backup, to copy the contents from the entire image chain, consisting of images [A], and [B], to the target image 'e.qcow2':
(QEMU) blockdev-backup device=node-B target=node-E sync=full job-id=job0
{
"execute": "blockdev-backup",
"arguments": {
"device": "node-B",
"job-id": "job0",
"target": "node-E",
"sync": "full"
}
}
Once the above 'backup' operation has completed, an event, BLOCK_JOB_COMPLETED, will be emitted, signalling successful completion.
Next, query for any active block device jobs (there should be none):
(QEMU) query-block-jobs
{
"execute": "query-block-jobs",
"arguments": {}
}
Shutdown the guest:
(QEMU) quit
{
"execute": "quit",
"arguments": {}
}
"return": {}
}
(QEMU)
{u'timestamp': {u'seconds': 1496072942, u'microseconds': 685292}, u'event': u'SHUTDOWN'}
Note
The above step is really important; if forgotten, an error, "Failed to get shared "write" lock on e.qcow2", will be thrown when you do qemu-img compare to verify the integrity of the backup disk image.
The end result will be, the image 'e.qcow2' containing a point-in-time backup of the disk image chain -- i.e. contents from images [A], and [B] at the time the blockdev-backup command was initiated.
One way to confirm the backup disk image contains the identical content with the disk image chain is to compare the backup, and the contents of the chain, you should see "Images are identical". (NB: this is assuming QEMU was launched with -S option, which will not start the CPUs at guest boot up):
$ qemu-img compare b.qcow2 e.qcow2 Warning: Image size mismatch! Images are identical.
NOTE: The "Warning: Image size mismatch!" is expected, as we created the target image (e.qcow2) with 39M size.