cinder/doc/source/admin/blockstorage-api-throughput.rst
Jay S. Bryant e9857d616d Migrate the blockstorage admin-guide to Cinder
This patch is part of the docs migration for Cinder.
It is more or less a drag and drop of the docs from
openstack-manuals admin-guide directory.  I needed to
change some syntax to work with Cinder's more stringent
doc build.

Note that the purpose of this patch is just to get the
documentation back to an accessible location.  Later patches
will clean up the organization and content.

Change-Id: Ib3f9255e0f9f2ff42a0ee4126607ff319a3d901e
2017-07-08 12:44:49 -05:00

35 lines
1.4 KiB
ReStructuredText

=============================================
Increase Block Storage API service throughput
=============================================
By default, the Block Storage API service runs in one process. This
limits the number of API requests that the Block Storage service can
process at any given time. In a production environment, you should
increase the Block Storage API throughput by allowing the Block Storage
API service to run in as many processes as the machine capacity allows.
.. note::
The Block Storage API service is named ``openstack-cinder-api`` on
the following distributions: CentOS, Fedora, openSUSE, Red Hat
Enterprise Linux, and SUSE Linux Enterprise. In Ubuntu and Debian
distributions, the Block Storage API service is named ``cinder-api``.
To do so, use the Block Storage API service option ``osapi_volume_workers``.
This option allows you to specify the number of API service workers
(or OS processes) to launch for the Block Storage API service.
To configure this option, open the ``/etc/cinder/cinder.conf``
configuration file and set the ``osapi_volume_workers`` configuration
key to the number of CPU cores/threads on a machine.
On distributions that include ``openstack-config``, you can configure
this by running the following command instead:
.. code-block:: console
# openstack-config --set /etc/cinder/cinder.conf \
DEFAULT osapi_volume_workers CORES
Replace ``CORES`` with the number of CPU cores/threads on a machine.