summaryrefslogtreecommitdiffstats
path: root/doc/dev/ceph-volume
diff options
context:
space:
mode:
Diffstat (limited to '')
-rw-r--r--doc/dev/ceph-volume/index.rst14
-rw-r--r--doc/dev/ceph-volume/lvm.rst179
-rw-r--r--doc/dev/ceph-volume/plugins.rst65
-rw-r--r--doc/dev/ceph-volume/systemd.rst37
-rw-r--r--doc/dev/ceph-volume/zfs.rst176
5 files changed, 471 insertions, 0 deletions
diff --git a/doc/dev/ceph-volume/index.rst b/doc/dev/ceph-volume/index.rst
new file mode 100644
index 000000000..5feef8089
--- /dev/null
+++ b/doc/dev/ceph-volume/index.rst
@@ -0,0 +1,14 @@
+===================================
+ceph-volume developer documentation
+===================================
+
+.. rubric:: Contents
+
+.. toctree::
+ :maxdepth: 1
+
+
+ plugins
+ lvm
+ zfs
+ systemd
diff --git a/doc/dev/ceph-volume/lvm.rst b/doc/dev/ceph-volume/lvm.rst
new file mode 100644
index 000000000..f2df6d850
--- /dev/null
+++ b/doc/dev/ceph-volume/lvm.rst
@@ -0,0 +1,179 @@
+
+.. _ceph-volume-lvm-api:
+
+LVM
+===
+The backend of ``ceph-volume lvm`` is LVM, it relies heavily on the usage of
+tags, which is a way for LVM to allow extending its volume metadata. These
+values can later be queried against devices and it is how they get discovered
+later.
+
+.. warning:: These APIs are not meant to be public, but are documented so that
+ it is clear what the tool is doing behind the scenes. Do not alter
+ any of these values.
+
+
+.. _ceph-volume-lvm-tag-api:
+
+Tag API
+-------
+The process of identifying logical volumes as part of Ceph relies on applying
+tags on all volumes. It follows a naming convention for the namespace that
+looks like::
+
+ ceph.<tag name>=<tag value>
+
+All tags are prefixed by the ``ceph`` keyword to claim ownership of that
+namespace and make it easily identifiable. This is how the OSD ID would be used
+in the context of lvm tags::
+
+ ceph.osd_id=0
+
+
+.. _ceph-volume-lvm-tags:
+
+Metadata
+--------
+The following describes all the metadata from Ceph OSDs that is stored on an
+LVM volume:
+
+
+``type``
+--------
+Describes if the device is an OSD or Journal, with the ability to expand to
+other types when supported (for example a lockbox)
+
+Example::
+
+ ceph.type=osd
+
+
+``cluster_fsid``
+----------------
+Example::
+
+ ceph.cluster_fsid=7146B649-AE00-4157-9F5D-1DBFF1D52C26
+
+
+``data_device``
+---------------
+Example::
+
+ ceph.data_device=/dev/ceph/data-0
+
+
+``data_uuid``
+-------------
+Example::
+
+ ceph.data_uuid=B76418EB-0024-401C-8955-AE6919D45CC3
+
+
+``journal_device``
+------------------
+Example::
+
+ ceph.journal_device=/dev/ceph/journal-0
+
+
+``journal_uuid``
+----------------
+Example::
+
+ ceph.journal_uuid=2070E121-C544-4F40-9571-0B7F35C6CB2B
+
+
+``encrypted``
+-------------
+Example for enabled encryption with ``luks``::
+
+ ceph.encrypted=1
+
+When encryption is not supported or simply disabled::
+
+ ceph.encrypted=0
+
+
+``osd_fsid``
+------------
+Example::
+
+ ceph.osd_fsid=88ab9018-f84b-4d62-90b4-ce7c076728ff
+
+
+``osd_id``
+----------
+Example::
+
+ ceph.osd_id=1
+
+
+``block_device``
+----------------
+Just used on :term:`bluestore` backends. Captures the path to the logical
+volume path.
+
+Example::
+
+ ceph.block_device=/dev/mapper/vg-block-0
+
+
+``block_uuid``
+--------------
+Just used on :term:`bluestore` backends. Captures either the logical volume UUID or
+the partition UUID.
+
+Example::
+
+ ceph.block_uuid=E5F041BB-AAD4-48A8-B3BF-31F7AFD7D73E
+
+
+``db_device``
+-------------
+Just used on :term:`bluestore` backends. Captures the path to the logical
+volume path.
+
+Example::
+
+ ceph.db_device=/dev/mapper/vg-db-0
+
+
+``db_uuid``
+-----------
+Just used on :term:`bluestore` backends. Captures either the logical volume UUID or
+the partition UUID.
+
+Example::
+
+ ceph.db_uuid=F9D02CF1-31AB-4910-90A3-6A6302375525
+
+
+``wal_device``
+--------------
+Just used on :term:`bluestore` backends. Captures the path to the logical
+volume path.
+
+Example::
+
+ ceph.wal_device=/dev/mapper/vg-wal-0
+
+
+``wal_uuid``
+------------
+Just used on :term:`bluestore` backends. Captures either the logical volume UUID or
+the partition UUID.
+
+Example::
+
+ ceph.wal_uuid=A58D1C68-0D6E-4CB3-8E99-B261AD47CC39
+
+
+``vdo``
+-------
+A VDO-enabled device is detected when device is getting prepared, and then
+stored for later checks when activating. This affects mount options by
+appending the ``discard`` mount flag, regardless of mount flags being used.
+
+Example for an enabled VDO device::
+
+ ceph.vdo=1
diff --git a/doc/dev/ceph-volume/plugins.rst b/doc/dev/ceph-volume/plugins.rst
new file mode 100644
index 000000000..95bc761e2
--- /dev/null
+++ b/doc/dev/ceph-volume/plugins.rst
@@ -0,0 +1,65 @@
+.. _ceph-volume-plugins:
+
+Plugins
+=======
+``ceph-volume`` started initially to provide support for using ``lvm`` as
+the underlying system for an OSD. It is included as part of the tool but it is
+treated like a plugin.
+
+This modularity, allows for other device or device-like technologies to be able
+to consume and re-use the utilities and workflows provided.
+
+Adding Plugins
+--------------
+As a Python tool, plugins ``setuptools`` entry points. For a new plugin to be
+available, it should have an entry similar to this in its ``setup.py`` file:
+
+.. code-block:: python
+
+ setup(
+ ...
+ entry_points = dict(
+ ceph_volume_handlers = [
+ 'my_command = my_package.my_module:MyClass',
+ ],
+ ),
+
+The ``MyClass`` should be a class that accepts ``sys.argv`` as its argument,
+``ceph-volume`` will pass that in at instantiation and call them ``main``
+method.
+
+This is how a plugin for ``ZFS`` could look like for example:
+
+.. code-block:: python
+
+ class ZFS(object):
+
+ help_menu = 'Deploy OSDs with ZFS'
+ _help = """
+ Use ZFS as the underlying technology for OSDs
+
+ --verbose Increase the verbosity level
+ """
+
+ def __init__(self, argv):
+ self.argv = argv
+
+ def main(self):
+ parser = argparse.ArgumentParser()
+ args = parser.parse_args(self.argv)
+ ...
+
+And its entry point (via ``setuptools``) in ``setup.py`` would looke like:
+
+.. code-block:: python
+
+ entry_points = {
+ 'ceph_volume_handlers': [
+ 'zfs = ceph_volume_zfs.zfs:ZFS',
+ ],
+ },
+
+After installation, the ``zfs`` subcommand would be listed and could be used
+as::
+
+ ceph-volume zfs
diff --git a/doc/dev/ceph-volume/systemd.rst b/doc/dev/ceph-volume/systemd.rst
new file mode 100644
index 000000000..8553430ee
--- /dev/null
+++ b/doc/dev/ceph-volume/systemd.rst
@@ -0,0 +1,37 @@
+.. _ceph-volume-systemd-api:
+
+systemd
+=======
+The workflow to *"activate"* an OSD is by relying on systemd unit files and its
+ability to persist information as a suffix to the instance name.
+
+``ceph-volume`` exposes the following convention for unit files::
+
+ ceph-volume@<sub command>-<extra metadata>
+
+For example, this is how enabling an OSD could look like for the
+:ref:`ceph-volume-lvm` sub command::
+
+ systemctl enable ceph-volume@lvm-0-8715BEB4-15C5-49DE-BA6F-401086EC7B41
+
+
+These 3 pieces of persisted information are needed by the sub-command so that
+it understands what OSD it needs to activate.
+
+Since ``lvm`` is not the only subcommand that will be supported, this
+is how it will allow other device types to be defined.
+
+At some point for example, for plain disks, it could be::
+
+ systemctl enable ceph-volume@disk-0-8715BEB4-15C5-49DE-BA6F-401086EC7B41
+
+At startup, the systemd unit will execute a helper script that will parse the
+suffix and will end up calling ``ceph-volume`` back. Using the previous
+example for lvm, that call will look like::
+
+ ceph-volume lvm activate 0 8715BEB4-15C5-49DE-BA6F-401086EC7B41
+
+
+.. warning:: These workflows are not meant to be public, but are documented so that
+ it is clear what the tool is doing behind the scenes. Do not alter
+ any of these values.
diff --git a/doc/dev/ceph-volume/zfs.rst b/doc/dev/ceph-volume/zfs.rst
new file mode 100644
index 000000000..18de7652a
--- /dev/null
+++ b/doc/dev/ceph-volume/zfs.rst
@@ -0,0 +1,176 @@
+
+.. _ceph-volume-zfs-api:
+
+ZFS
+===
+The backend of ``ceph-volume zfs`` is ZFS, it relies heavily on the usage of
+tags, which is a way for ZFS to allow extending its volume metadata. These
+values can later be queried against devices and it is how they get discovered
+later.
+
+Currently this interface is only usable when running on FreeBSD.
+
+.. warning:: These APIs are not meant to be public, but are documented so that
+ it is clear what the tool is doing behind the scenes. Do not alter
+ any of these values.
+
+
+.. _ceph-volume-zfs-tag-api:
+
+Tag API
+-------
+The process of identifying filesystems, volumes and pools as part of Ceph relies
+on applying tags on all volumes. It follows a naming convention for the
+namespace that looks like::
+
+ ceph.<tag name>=<tag value>
+
+All tags are prefixed by the ``ceph`` keyword to claim ownership of that
+namespace and make it easily identifiable. This is how the OSD ID would be used
+in the context of zfs tags::
+
+ ceph.osd_id=0
+
+Tags on filesystems are stored as property.
+Tags on a zpool are stored in the comment property as a concatenated list
+separated by ``;``
+
+.. _ceph-volume-zfs-tags:
+
+Metadata
+--------
+The following describes all the metadata from Ceph OSDs that is stored on a
+ZFS filesystem, volume, pool:
+
+
+``type``
+--------
+Describes if the device is an OSD or Journal, with the ability to expand to
+other types when supported
+
+Example::
+
+ ceph.type=osd
+
+
+``cluster_fsid``
+----------------
+Example::
+
+ ceph.cluster_fsid=7146B649-AE00-4157-9F5D-1DBFF1D52C26
+
+
+``data_device``
+---------------
+Example::
+
+ ceph.data_device=/dev/ceph/data-0
+
+
+``data_uuid``
+-------------
+Example::
+
+ ceph.data_uuid=B76418EB-0024-401C-8955-AE6919D45CC3
+
+
+``journal_device``
+------------------
+Example::
+
+ ceph.journal_device=/dev/ceph/journal-0
+
+
+``journal_uuid``
+----------------
+Example::
+
+ ceph.journal_uuid=2070E121-C544-4F40-9571-0B7F35C6CB2B
+
+
+``osd_fsid``
+------------
+Example::
+
+ ceph.osd_fsid=88ab9018-f84b-4d62-90b4-ce7c076728ff
+
+
+``osd_id``
+----------
+Example::
+
+ ceph.osd_id=1
+
+
+``block_device``
+----------------
+Just used on :term:`bluestore` backends. Captures the path to the logical
+volume path.
+
+Example::
+
+ ceph.block_device=/dev/gpt/block-0
+
+
+``block_uuid``
+--------------
+Just used on :term:`bluestore` backends. Captures either the logical volume UUID or
+the partition UUID.
+
+Example::
+
+ ceph.block_uuid=E5F041BB-AAD4-48A8-B3BF-31F7AFD7D73E
+
+
+``db_device``
+-------------
+Just used on :term:`bluestore` backends. Captures the path to the logical
+volume path.
+
+Example::
+
+ ceph.db_device=/dev/gpt/db-0
+
+
+``db_uuid``
+-----------
+Just used on :term:`bluestore` backends. Captures either the logical volume UUID or
+the partition UUID.
+
+Example::
+
+ ceph.db_uuid=F9D02CF1-31AB-4910-90A3-6A6302375525
+
+
+``wal_device``
+--------------
+Just used on :term:`bluestore` backends. Captures the path to the logical
+volume path.
+
+Example::
+
+ ceph.wal_device=/dev/gpt/wal-0
+
+
+``wal_uuid``
+------------
+Just used on :term:`bluestore` backends. Captures either the logical volume UUID or
+the partition UUID.
+
+Example::
+
+ ceph.wal_uuid=A58D1C68-0D6E-4CB3-8E99-B261AD47CC39
+
+
+``compression``
+---------------
+A compression-enabled device can always be set using the native zfs settings on
+a volume or filesystem. This will/can be activated during creation of the volume
+of filesystem.
+When activated by ``ceph-volume zfs`` this tag will be created.
+Compression manually set AFTER ``ceph-volume`` will go unnoticed, unless this
+tag is also manually set.
+
+Example for an enabled compression device::
+
+ ceph.vdo=1