9:04 a.m.
On Mon, Jul 08, 2019 at 11:55:45 -0500, Eric Blake wrote:
Prepare for new backup APIs by describing the XML that will
represent
a backup. The XML resembles snapshots and checkpoints in being able
to select actions for a set of disks, but has other differences. It
can support both push model (the hypervisor does the backup directly
into the destination file) and pull model (the hypervisor exposes an
access port for a third party to grab what is necessary). Add
testsuite coverage for some minimal uses of the XML.
The <disk> element within <domainbackup> tries to model the same
elements as a <disk> under <domain>, but sharing the RNG grammar
proved to be hairy. That is in part because while <domain> use
<source> to describe a host resource in use by the guest, a backup job
is using a host resource that is not visible to the guest: a push
backup action is instead describing a <target> (which ultimately could
be a remote network resource, but for simplicity the RNG just
validates a local file for now), and a pull backup action is instead
describing a temporary local file <scratch> (which probably should not
be a remote resource). A future refactoring may thus introduce some
way to parameterize RNG to accept <disk type='FOO'>...</disk> so
that
the name of the subelement can be <source> for domain, or <target> or
<scratch> as needed for backups. Future patches may improve this area
of code.
Signed-off-by: Eric Blake <eblake(a)redhat.com>
---
[...]
diff --git a/docs/formatbackup.html.in b/docs/formatbackup.html.in
new file mode 100644
index 0000000000..4f76013d84
--- /dev/null
+++ b/docs/formatbackup.html.in
@@ -0,0 +1,184 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE html>
+<html xmlns="http://www.w3.org/1999/xhtml">
+ <body>
+ <h1>Backup XML format</h1>
+
+ <ul id="toc"></ul>
+
+ <h2><a id="BackupAttributes">Backup XML</a></h2>
+
+ <p>
+ Creating a backup, whether full or incremental, is done
+ via <code>virDomainBackupBegin()</code>, which takes an XML
+ description of the actions to perform, as well as an optional
+ second XML document <a href="formatcheckpoint.html">describing a
+ checkpoint</a> to create at the same point in time. See
+ also <a href="domainstatecapture.html">a comparison</a>
between
+ the various state capture APIs.
+ </p>
+ <p>
+ There are two general modes for backups: a push mode (where the
+ hypervisor writes out the data to the destination file, which
+ may be local or remote), and a pull mode (where the hypervisor
+ creates an NBD server that a third-party client can then read as
+ needed, and which requires the use of temporary storage,
+ typically local, until the backup is complete).
+ </p>
+ <p>
+ The instructions for beginning a backup job are provided as
+ attributes and elements of the
+ top-level <code>domainbackup</code> element. This element
+ includes an optional attribute <code>mode</code> which can be
+ either "push" or "pull" (default
+ push). <code>virDomainBackupGetXMLDesc()</code> can be used to
+ see the actual values selected for elements omitted during
+ creation (for example, learning which port the NBD server is
+ using in the pull model or what file names libvirt generated
+ when none were supplied). The following child elements are
+ supported:
+ </p>
+ <dl>
+ <dt><code>id</code></dt>
+ <dd>Ignored on input, this element is the job id of the backup
+ operation returned on success
+ from <code>virDomainBackupBegin()</code>, and is used for
+ selecting which backup operation to target
+ during <code>virDomainBackupGetXMLDesc()</code>
Rather than adding yet another job-specific API I'm thinking of adding a
set of "domain job" APIs which will allow getting XML stats for
migration block and other jobs generically.
My motivation is that apart from this, we also have a possibility to
have a blockjob which does not have a parent disk any more (e.g. because
the guest ejected it) and thus the current APIs would not semantically
work with such a thing.
By having a generic API we can fix all of these problems without more
single-use APIs.
+ and <code>virDomainBackupEnd()</code>. (Note
that until
The same for this one. Obviously the job identifier will then become a
string rather than a number which will also encode the job type e.g.:
block-pull-NODENAME or block-backup-3 etc.
I'll try to send a example tomorrow or so. (I didn't start yet though)
+ additional APIs are added for supporting parallel jobs, it
is
+ also possible to ignore this element and use the job
+ id <code>0</code> to refer to the one and only current backup
+ job.)</dd>
+ <dt><code>incremental</code></dt>
+ <dd>An optional element giving the name of an existing
+ checkpoint of the domain, which will be used to make this
+ backup an incremental one. In the push model, only changes
+ since the named checkpoint are written to the destination. In
+ the pull model, the NBD server uses the
+ NBD_OPT_SET_META_CONTEXT extension to advertise to the client
+ which portions of the export contain changes since the named
+ checkpoint. If omitted, a full backup is performed.
+ </dd>
+ <dt><code>server</code></dt>
+ <dd>Present only for a pull mode backup. Contains the same
+ attributes as
+ the <a
href="formatdomain.html#elementsDisks"><code>protocol</code>
+ element of a disk</a> attached via NBD in the domain (such as
So will it also contain the 'procotol' setting itself? For future
proofing we should not mandate NBD only in the XML.
+ transport, socket, name, port, or tls), necessary to set up
an
+ NBD server that exposes the content of each disk at the time
+ the backup is started.
+ </dd>
+ <dt><code>disks</code></dt>
+ <dd>An optional listing of instructions for disks participating
+ in the backup (if omitted, all disks participate and libvirt
+ attempts to generate filenames by appending the current
+ timestamp as a suffix). If the entire element was omitted on
+ input, then all disks participate in the backup, otherwise,
+ only the disks explicitly listed which do not also
+ use <code>backup='no'</code> will participate. On output,
this
+ is the state of each of the domain's disk in relation to the
+ backup operation.
+ <dl>
+ <dt><code>disk</code></dt>
+ <dd>This sub-element describes the backup properties of a
+ specific disk, with the following attributes and child
+ elements:
+ <dl>
+ <dt><code>name</code></dt>
+ <dd>A mandatory attribute which must match either
+ the <code><target dev='name'/></code>
or an
+ unambiguous <code><source
file='name'/></code>
+ of one of
+ the <a href="formatdomain.html#elementsDisks">disk
+ devices</a> specified for the domain at the time of
+ the checkpoint.</dd>
+ <dt><code>backup</code></dt>
+ <dd>An optional attribute to describe the state of the
+ backup for the given disk. On input, the
+ value <code>no</code> skips a backup of the disk, and
+ any other value is ignored; on output, other values
+ such as <code>begin</code>,
<code>inprogress</code>,
+ or <code>ready</code> track the backup progress that
Hmmm, I think the input and output XML descriptions should be separated,
it's getting quite confusing.
+ libvirt has observed for that disk.</dd>
+ <dt><code>type</code></dt>
+ <dd>An optional attribute to describe the type of the
+ disk, except when <code>backup='no'</code> is
+ used. Valid values include <code>file</code>
+ or <code>block</code> for both push and pull model
+ backups; in the future, <code>network</code> may be
+ added for push model backups. Similar to a disk
+ declaration for a domain, the choice of type controls
+ what additional sub-elements are needed to describe
+ the destination (such as <code>protocol</code> for a
+ network destination).</dd>
+ <dt><code>target</code></dt>
+ <dd>Valid only for push mode backups, this is the
+ primary sub-element that describes the file name of
+ the backup destination, similar to
+ the <code>source</code> sub-element of a domain
+ disk. An optional sub-element <code>driver</code> can
+ also be used, with an attribute <code>type</code> to
+ specify a destination format different from
+ qcow2. Additionally, if a push backup is not
+ incremental, <code>target</code> may contain an
+ optional attribute <code>shallow="on"</code> so
that
+ the destination file copies only the top-most source
+ file in a backing chain, rather than collapsing the
+ entire chain into the copy.</dd>
+ <dt><code>scratch</code></dt>
+ <dd>Valid only for pull mode backups, this is the
+ primary sub-element that describes the file name of
+ the local scratch file to be used in facilitating the
+ backup, and is similar to the <code>source</code>
The source subelement does not contain the 'type' attribute. Where do we
set it here? <scratch type='block'><source dev='/dev/blah'/>
?
+ sub-element of a domain disk.</dd>
+ </dl>
+ </dd>
+ </dl>
+ </dd>
+ </dl>
+
+ <h2><a id="example">Examples</a></h2>
+
+ <p>Use <code>virDomainBackupBegin()</code> to perform a full
+ backup using push mode. The example lets libvirt pick the
+ destination and format for 'vda', fully specifies that we want a
+ raw backup of 'vdb', and omits 'vdc' from the operation.
+ </p>
+ <pre>
+<domainbackup>
+ <disks/>
+ <disk name='vda'/>
+ <disk name='vdb' type='file'>
+ <target file='/path/to/vdb.backup'/>
+ <driver type='raw'/>
+ </disk>
+ <disk name='vdc' backup='no'/>
+ </disks/>
+</domainbackup>
+ </pre>
+
+ <p>If the previous full backup also passed a parameter describing
+ <a href="formatcheckpoint.html">checkpoint XML</a> that
resulted
+ in a checkpoint named <code>1525889631</code>, we can make
+ another call to <code>virDomainBackupBegin()</code> to perform
+ an incremental backup of just the data changed since that
+ checkpoint, this time using the following XML to start a pull
+ model export of the 'vda' and 'vdb' disks, where a third-party
+ NBD client connecting to '/path/to/server' completes the backup
+ (omitting 'vdc' from the explicit list has the same effect as
+ the backup='no' from the previous example):
+ </p>
+ <pre>
+<domainbackup mode="pull">
+ <incremental>1525889631</incremental>
+ <server transport="unix" socket="/path/to/server"/>
+ <disks/>
+ <disk name='vda' type='file'>
+ <scratch file='/path/to/file1.scratch'/>
I think you want to be able to describe the type somehow. I'm expecting
RHV folks to want to stash this into an LV any second.
+ </disk>
+ </disks/>
+</domainbackup>
+ </pre>
+ </body>
+</html>
[...]
diff --git a/docs/schemas/domainbackup.rng
b/docs/schemas/domainbackup.rng
new file mode 100644
index 0000000000..92327e7077
--- /dev/null
+++ b/docs/schemas/domainbackup.rng
@@ -0,0 +1,219 @@
+<?xml version="1.0"?>
+<!-- A Relax NG schema for the libvirt domain backup properties XML format -->
+<grammar xmlns="http://relaxng.org/ns/structure/1.0">
+ <start>
+ <ref name='domainbackup'/>
+ </start>
+
+ <include href='domaincommon.rng'/>
+
+ <define name='domainbackup'>
+ <element name='domainbackup'>
+ <optional>
+ <attribute name='id'>
+ <ref name="unsignedInt"/>
+ </attribute>
+ </optional>
+ <interleave>
+ <optional>
+ <element name='incremental'>
+ <text/>
name
+ </element>
+ </optional>
+ <choice>
+ <group>
+ <optional>
+ <attribute name='mode'>
+ <value>push</value>
+ </attribute>
+ </optional>
+ <ref name='backupDisksPush'/>
+ </group>
+ <group>
+ <attribute name='mode'>
+ <value>pull</value>
+ </attribute>
+ <interleave>
+ <element name='server'>
+ <choice>
lacking 'protocol' field
+ <group>
+ <optional>
+ <attribute name='transport'>
+ <value>tcp</value>
+ </attribute>
+ </optional>
+ <attribute name='name'>
+ <choice>
+ <ref name='dnsName'/>
+ <ref name='ipAddr'/>
+ </choice>
+ </attribute>
+ <optional>
+ <attribute name='port'>
+ <ref name='unsignedInt'/>
+ </attribute>
+ </optional>
+ <!-- add tls? -->
In 2019 the more appropriate question would be whether to allow non-tls
mode at all.
+ </group>
+ <group>
+ <attribute name='transport'>
+ <value>unix</value>
+ </attribute>
+ <attribute name='socket'>
+ <ref name='absFilePath'/>
+ </attribute>
+ </group>
+ </choice>
+ </element>
+ <ref name='backupDisksPull'/>
+ </interleave>
+ </group>
+ </choice>
+ </interleave>
+ </element>
+ </define>
+
+ <define name='backupPushDriver'>
+ <optional>
+ <element name='driver'>
+ <attribute name='type'>
+ <ref name='storageFormat'/>
+ </attribute>
+ </element>
+ </optional>
+ </define>
+
+ <define name='backupAttr'>
+ <!-- valid values of <disk backup='XXX'> other than 'no'
-->
+ <optional>
+ <attribute name='backup'>
+ <choice>
+ <value>begin</value>
+ <value>inprogress</value>
+ <value>ready</value>
+ </choice>
+ </attribute>
+ </optional>
+ </define>
+
+ <define name='backupDisksPush'>
+ <optional>
+ <element name='disks'>
+ <oneOrMore>
+ <element name='disk'>
+ <attribute name='name'>
+ <choice>
+ <ref name='diskTarget'/>
+ <ref name='absFilePath'/>
+ </choice>
+ </attribute>
+ <optional>
+ <attribute name='shallow'>
+ <value>on</value>
+ </attribute>
+ </optional>
+ <choice>
+ <group>
+ <attribute name='backup'>
+ <value>no</value>
+ </attribute>
+ </group>
+ <!-- FIXME allow push to a network location, perhaps by
+ refactoring 'diskSource' to take element name by a
+ per-grammar ref -->
+ <group>
+ <ref name='backupAttr'/>
+ <optional>
+ <attribute name='type'>
+ <value>file</value>
+ </attribute>
+ </optional>
+ <interleave>
+ <optional>
+ <element name='target'>
+ <attribute name='file'>
+ <ref name='absFilePath'/>
+ </attribute>
+ </element>
+ </optional>
+ <ref name='backupPushDriver'/>
+ </interleave>
+ </group>
+ <group>
+ <ref name='backupAttr'/>
+ <attribute name='type'>
+ <value>block</value>
+ </attribute>
+ <interleave>
+ <optional>
+ <element name='target'>
+ <attribute name='dev'>
+ <ref name='absFilePath'/>
+ </attribute>
+ </element>
+ </optional>
+ <ref name='backupPushDriver'/>
+ </interleave>
+ </group>
+ <!--
+ <group>
+ <ref name='backupAttr'/>
+ <attribute name='type'>
+ <value>network</value>
+ </attribute>
+ ...
+ </group>
+ -->
+ </choice>
+ </element>
+ </oneOrMore>
+ </element>
+ </optional>
+ </define>
+
+ <define name='backupDisksPull'>
+ <optional>
+ <element name='disks'>
+ <oneOrMore>
+ <element name='disk'>
+ <attribute name='name'>
+ <choice>
+ <ref name='diskTarget'/>
+ <ref name='absFilePath'/>
+ </choice>
+ </attribute>
+ <choice>
+ <group>
+ <optional>
Interresting, there is a scratch location 'type'. So this really should
not be optional.
+ <attribute name='type'>
+ <value>file</value>
+ </attribute>
+ </optional>
+ <optional>
+ <element name='scratch'>
+ <attribute name='file'>
+ <ref name='absFilePath'/>
+ </attribute>
+ </element>
+ </optional>
+ </group>
+ <group>
+ <attribute name='type'>
+ <value>block</value>
also mention this in the example/test?
+ </attribute>
+ <optional>
+ <element name='scratch'>
+ <attribute name='dev'>
+ <ref name='absFilePath'/>
+ </attribute>
+ </element>
+ </optional>
+ </group>
+ </choice>
+ </element>
+ </oneOrMore>
+ </element>
+ </optional>
+ </define>
+
+</grammar>
diff --git a/tests/domainbackupxml2xmlin/backup-pull.xml
b/tests/domainbackupxml2xmlin/backup-pull.xml
new file mode 100644
index 0000000000..2ce5cd6711
--- /dev/null
+++ b/tests/domainbackupxml2xmlin/backup-pull.xml
@@ -0,0 +1,9 @@
+<domainbackup mode="pull">
+ <incremental>1525889631</incremental>
+ <server transport='tcp' name='localhost' port='10809'/>
This really needs a 'protocol' element and also potentially multiple
host elements.
+ <disks>
+ <disk name='vda' type='file'>
+ <scratch file='/path/to/file'/>
and this really needs something else than a file for the scratch
+ </disk>
+ </disks>
+</domainbackup>
[...]