From c719ab2d6bc8dec5c43ab5a536789f26e97b1db2 Mon Sep 17 00:00:00 2001 From: KATO Tomoyuki Date: Wed, 2 Dec 2015 22:54:47 +0900 Subject: [PATCH] [config-ref] Convert compute fc and iscsi to RST Change-Id: Idc4209f7217ff4b2bbbe2eeeb9a978a18dcbac30 Implements: blueprint config-ref-rst --- doc/config-ref-rst/source/compute.rst | 2 + .../source/compute/fibre-channel.rst | 26 +++++++ .../source/compute/iscsi-offload.rst | 76 +++++++++++++++++++ 3 files changed, 104 insertions(+) create mode 100644 doc/config-ref-rst/source/compute/fibre-channel.rst create mode 100644 doc/config-ref-rst/source/compute/iscsi-offload.rst diff --git a/doc/config-ref-rst/source/compute.rst b/doc/config-ref-rst/source/compute.rst index f52ee45b32..58e37e3597 100644 --- a/doc/config-ref-rst/source/compute.rst +++ b/doc/config-ref-rst/source/compute.rst @@ -13,6 +13,8 @@ Compute compute/rpc.rst compute/api.rst compute/ec2-api.rst + compute/fibre-channel.rst + compute/iscsi-offload.rst compute/hypervisors.rst compute/scheduler.rst compute/cells.rst diff --git a/doc/config-ref-rst/source/compute/fibre-channel.rst b/doc/config-ref-rst/source/compute/fibre-channel.rst new file mode 100644 index 0000000000..eec82030cd --- /dev/null +++ b/doc/config-ref-rst/source/compute/fibre-channel.rst @@ -0,0 +1,26 @@ +================================ +Fibre Channel support in Compute +================================ + +Fibre Channel support in OpenStack Compute is remote block +storage attached to compute nodes for VMs. + +.. TODO: This below statement needs to be verified for current release + +Fibre Channel supported only the KVM hypervisor. + +Compute and Block Storage support Fibre Channel automatic zoning on +Brocade and Cisco switches. On other hardware Fibre Channel arrays must +be pre-zoned or directly attached to the KVM hosts. + +KVM host requirements +~~~~~~~~~~~~~~~~~~~~~ + +You must install these packages on the KVM host: + +* ``sysfsutils`` - Nova uses the ``systool`` application in this package. +* ``sg3-utils`` or ``sg3_utils`` - Nova uses the ``sg_scan`` and + ``sginfo`` applications. + +Installing the ``multipath-tools`` or ``device-mapper-multipath`` +package is optional. diff --git a/doc/config-ref-rst/source/compute/iscsi-offload.rst b/doc/config-ref-rst/source/compute/iscsi-offload.rst new file mode 100644 index 0000000000..2602a5b706 --- /dev/null +++ b/doc/config-ref-rst/source/compute/iscsi-offload.rst @@ -0,0 +1,76 @@ +============================================== +iSCSI interface and offload support in Compute +============================================== + +.. note:: + + iSCSI interface and offload support is only present since Kilo. + +Compute supports open-iscsi iSCSI interfaces for offload cards. +Offload hardware must be present and configured on every compute +node where offload is desired. Once an open-iscsi interface is +configured, the iface name (``iface.iscsi_ifacename``) should be +passed to libvirt via the ``iscsi_iface`` parameter for use. +All iSCSI sessions will be bound to this iSCSI interface. + +Currently supported transports (``iface.transport_name``) are +``be2iscsi``, ``bnx2i``, ``cxgb3i``, ``cxgb4i``, ``qla4xxx``, ``ocs``. +Configuration changes are required on the compute node only. + +iSER is supported using the separate iSER LibvirtISERVolumeDriver +and will be rejected if used via the iscsi_iface parameter. + +iSCSI iface configuration +~~~~~~~~~~~~~~~~~~~~~~~~~ + +* Note the distinction between the transport name (``iface.transport_name``) + and iface name (``iface.iscsi_ifacename``). The actual iface name must be + specified via the iscsi_iface parameter to libvirt for offload to work. +* The default name for an iSCSI iface (open-iscsi parameter + ``iface.iscsi_ifacename``) is in the format transport_name.hwaddress + when generated by ``iscsiadm``. +* ``iscsiadm`` can be used to view and generate current iface configuration. + Every network interface that supports an open-iscsi transport can have one + or more iscsi ifaces associated with it. If no ifaces have been configured + for a network interface supported by an open-iscsi transport, + this command will create a default iface configuration for that + network interface. For example : + + .. code-block:: console + + # iscsiadm -m iface + default tcp,,,, + iser iser,,,, + bnx2i.00:05:b5:d2:a0:c2 bnx2i,00:05:b5:d2:a0:c2,5.10.10.20,, + + The output is in the format: + ``iface_name transport_name,hwaddress,ipaddress, + net_ifacename,initiatorname``. + +* Individual iface configuration can be viewed via + + .. code-block:: console + + # iscsiadm -m iface -I IFACE_NAME + # BEGIN RECORD 2.0-873 + iface.iscsi_ifacename = cxgb4i.00:07:43:28:b2:58 + iface.net_ifacename = + iface.ipaddress = 102.50.50.80 + iface.hwaddress = 00:07:43:28:b2:58 + iface.transport_name = cxgb4i + iface.initiatorname = + # END RECORD + + Configuration can be updated as desired via + + .. code-block:: console + + # iscsiadm -m iface-I IFACE_NAME--op=update -n iface.SETTING -v VALUE + +* All iface configurations need a minimum of ``iface.iface_name``, + ``iface.transport_name`` and ``iface.hwaddress`` to be correctly + configured to work. Some transports may require ``iface.ipaddress`` + and ``iface.net_ifacename`` as well to bind correctly. + + Detailed configuration instructions can be found at + http://www.open-iscsi.org/docs/README.