openstack-manuals/doc/config-reference/compute/section_compute-cells.xml
Gauvain Pocentek 590bb27b37 update nova flags and tables
Removed docker references (the driver has been removed from nova).

Update confilcting pages including tables to use xrefs instead.

Add a rpc_all section to handle conflicts in the training-guides.

Closes-Bug: #1291138
Closes-Bug: #1292612
Change-Id: I67ad31c3913cae4c2e9e3d2ce41302717af5d839
2014-03-23 07:41:52 +01:00

261 lines
16 KiB
XML

<section xml:id="section_compute-cells"
xmlns="http://docbook.org/ns/docbook"
xmlns:xlink="http://www.w3.org/1999/xlink"
xmlns:xi="http://www.w3.org/2001/XInclude"
xmlns:ns5="http://www.w3.org/1999/xhtml"
xmlns:ns4="http://www.w3.org/2000/svg"
xmlns:ns3="http://www.w3.org/1998/Math/MathML"
xmlns:ns="http://docbook.org/ns/docbook"
version="5.0">
<title>Cells</title>
<?dbhtml stop-chunking?>
<para><emphasis role="italic">Cells</emphasis> functionality allows you to scale an OpenStack
Compute cloud in a more distributed fashion without having to use complicated technologies
like database and message queue clustering. It is intended to support very large
deployments.</para>
<para>When this functionality is enabled, the hosts in an OpenStack Compute cloud are
partitioned into groups called cells. Cells are configured as a tree. The top-level cell
should have a host that runs a <systemitem class="service">nova-api</systemitem> service,
but no <systemitem class="service">nova-compute</systemitem> services. Each child cell
should run all of the typical <systemitem class="service">nova-*</systemitem> services in a
regular Compute cloud except for <systemitem class="service">nova-api</systemitem>. You can
think of cells as a normal Compute deployment in that each cell has its own database server
and message queue broker.</para>
<para>The <systemitem class="service">nova-cells</systemitem> service handles communication
between cells and selects cells for new instances. This service is required for every cell.
Communication between cells is pluggable, and currently the only option is communication through
RPC.</para>
<para>Cells scheduling is separate from host scheduling. <systemitem class="service">nova-cells</systemitem> first picks a
cell (now randomly, but future releases plan to add filtering/weighing functionality, and
decisions will be based on broadcasts of capacity/capabilities). Once a cell is
selected and the new build request reaches its <systemitem class="service">nova-cells</systemitem> service, it is sent
over to the host scheduler in that cell and the build proceeds as it would have without
cells.</para>
<warning>
<para>Cell functionality is currently considered experimental.</para>
</warning>
<section xml:id="cell-config-options">
<title>Cell configuration options</title>
<para>Cells are disabled by default. All cell-related configuration options go under a
<literal>[cells]</literal> section in <filename>nova.conf</filename>. The following
cell-related options are currently supported:<variablelist>
<varlistentry>
<term><literal>enable</literal></term>
<listitem>
<para>Set this is <literal>True</literal> to turn on cell functionality,
which is off by default.</para>
</listitem>
</varlistentry>
<varlistentry>
<term><literal>name</literal></term>
<listitem>
<para>Name of the current cell. This must be unique for each cell.</para>
</listitem>
</varlistentry>
<varlistentry>
<term><literal>capabilities</literal></term>
<listitem>
<para>List of arbitrary
<literal><replaceable>key</replaceable>=<replaceable>value</replaceable></literal>
pairs defining capabilities of the current cell.
Values include
<literal>hypervisor=xenserver;kvm,os=linux;windows</literal>.</para>
</listitem>
</varlistentry>
<varlistentry>
<term><literal>call_timeout</literal></term>
<listitem>
<para>How long in seconds to wait for replies from calls
between cells.</para>
</listitem>
</varlistentry>
<varlistentry>
<term>scheduler_filter_classes</term>
<listitem>
<para>Filter classes that the cells scheduler should use. By default, uses
"<literal>nova.cells.filters.all_filters</literal>" to map to all
cells filters included with Compute.</para>
</listitem>
</varlistentry>
<varlistentry>
<term>scheduler_weight_classes</term>
<listitem><para>Weight classes the cells scheduler should use. By default, uses
"<literal>nova.cells.weights.all_weighers</literal>"
to map to all cells weight algorithms (weighers)
included with Compute.</para></listitem>
</varlistentry>
<varlistentry>
<term>ram_weight_multiplier</term>
<listitem>
<para>Multiplier used for weighing ram. Negative numbers
mean you want Compute to stack VMs on one host
instead of spreading out new VMs to more hosts in
the cell. Default value is 10.0.</para>
</listitem>
</varlistentry>
</variablelist></para>
</section>
<section xml:id="config-API-cell">
<title>Configure the API (top-level) cell</title>
<para>The compute API class must be changed in the API cell so that requests can be proxied
through nova-cells down to the correct cell properly. Add the following to
<filename>nova.conf</filename> in the API
cell:<programlisting language="ini">[DEFAULT]
compute_api_class=nova.compute.cells_api.ComputeCellsAPI
...
[cells]
enable=True
name=api</programlisting></para>
</section>
<section xml:id="config-child-cell">
<title>Configure the child cells</title>
<para>Add the following to <filename>nova.conf</filename> in the child cells, replacing
<replaceable>cell1</replaceable> with the name of each
cell:<programlisting language="ini">[DEFAULT]
# Disable quota checking in child cells. Let API cell do it exclusively.
quota_driver=nova.quota.NoopQuotaDriver
[cells]
enable=True
name=<replaceable>cell1</replaceable></programlisting></para>
</section>
<section xml:id="config-cell-db">
<title>Configure the database in each cell</title>
<para>Before bringing the services online, the database in each cell needs to be configured
with information about related cells. In particular, the API cell needs to know about
its immediate children, and the child cells must know about their immediate agents.
The information needed is the <application>RabbitMQ</application> server credentials
for the particular cell.</para>
<para>Use the <command>nova-manage cell create</command> command to add this information to
the database in each
cell:<screen><prompt>#</prompt> <userinput>nova-manage cell create -h</userinput>
<computeroutput>Options:
-h, --help show this help message and exit
--name=&lt;name> Name for the new cell
--cell_type=&lt;parent|child>
Whether the cell is a parent or child
--username=&lt;username>
Username for the message broker in this cell
--password=&lt;password>
Password for the message broker in this cell
--hostname=&lt;hostname>
Address of the message broker in this cell
--port=&lt;number> Port number of the message broker in this cell
--virtual_host=&lt;virtual_host>
The virtual host of the message broker in this cell
--woffset=&lt;float>
(weight offset) It might be used by some cell scheduling code in the future
--wscale=&lt;float>
(weight scale) It might be used by some cell scheduling code in the future
</computeroutput></screen></para>
<para>As an example, assume we have an API cell named <literal>api</literal> and a child
cell named <literal>cell1</literal>. Within the api cell, we have the following RabbitMQ
server
info:<programlisting language="ini">rabbit_host=10.0.0.10
rabbit_port=5672
rabbit_username=api_user
rabbit_password=api_passwd
rabbit_virtual_host=api_vhost</programlisting></para>
<para>And in the child cell named <literal>cell1</literal> we have the following RabbitMQ
server
info:<programlisting language="ini">rabbit_host=10.0.1.10
rabbit_port=5673
rabbit_username=cell1_user
rabbit_password=cell1_passwd
rabbit_virtual_host=cell1_vhost</programlisting></para>
<para>We would run this in the API cell, as
root.<screen><prompt>#</prompt> <userinput>nova-manage cell create --name=cell1 --cell_type=child --username=cell1_user --password=cell1_passwd --hostname=10.0.1.10 --port=5673 --virtual_host=cell1_vhost --woffset=1.0 --wscale=1.0</userinput></screen></para>
<para>Repeat the above for all child cells.</para>
<para>In the child cell, we would run the following, as
root:<screen><prompt>#</prompt> <userinput>nova-manage cell create --name=api --cell_type=parent --username=api1_user --password=api1_passwd --hostname=10.0.0.10 --port=5672 --virtual_host=api_vhost --woffset=1.0 --wscale=1.0
</userinput></screen></para>
<para>To customize the Compute cells, use the configuration option
settings documented in <xref
linkend="config_table_nova_cells"/>.</para>
</section>
<section xml:id="cell-weights-filters">
<title>Cell scheduling configuration</title>
<para>To determine the best cell for launching a new instance, Compute uses a set of
filters and weights configured in <filename>/etc/nova/nova.conf</filename>. The
following options are available to prioritize cells for scheduling:</para>
<para>
<itemizedlist>
<listitem>
<para><code>scheduler_filter_classes</code> - Specifies the list of filter
classes. By default <code>nova.cells.weights.all_filters</code> is
specified, which maps to all cells filters included with Compute (see
<xref linkend="scheduler-filters"/>.</para>
</listitem>
<listitem>
<para><code>scheduler_weight_classes</code> - Specifies the list of weight
classes. By default <code>nova.cells.weights.all_weighers</code> is
specified, which maps to all cell weight algorithms (weighers) included with
Compute. The following modules are available:</para>
<itemizedlist><listitem>
<para><code>mute_child</code>: Downgrades the likelihood of child cells
being chosen for scheduling requests, which haven't sent capacity or
capability updates in a while. Options include
<code>mute_weight_multiplier</code> (multiplier for mute
children; value should be negative) and
<code>mute_weight_value</code> (assigned to mute children;
should be a positive value).</para>
</listitem>
<listitem>
<para><code>ram_by_instance_type</code>: Select cells with the most RAM
capacity for the instance type being requested. Because higher
weights win, Compute returns the number of available units for the
instance type requested. The <code>ram_weight_multiplier</code>
option defaults to 10.0 that adds to the weight by a factor of 10.
Use a negative number to stack VMs on one host instead of spreading
out new VMs to more hosts in the cell.</para>
</listitem>
<listitem><para><code>weight_offset</code>: Allows modifying the database to weight a particular cell. You can
use this when you want to disable a cell (for example, '0'), or to
set a default cell by making its weight_offset very high (for
example, '999999999999999'). The highest weight will be the first
cell to be scheduled for launching an instance.</para></listitem></itemizedlist>
</listitem>
</itemizedlist>
</para>
<para>Additionally, the following options are available for the cell scheduler:</para>
<para>
<itemizedlist>
<listitem>
<para><code>scheduler_retries</code> - Specifies how many times the scheduler
tries to launch a new instance when no cells are available (default=10).</para>
</listitem>
<listitem>
<para><code>scheduler_retry_delay</code> - Specifies the delay (in seconds)
between retries (default=2).</para>
</listitem>
</itemizedlist>
</para>
<para>As an admin user, you can also add a filter that directs builds to
a particular cell. The <filename>policy.json</filename> file must
have a line with <literal>"cells_scheduler_filter:TargetCellFilter"
: "is_admin:True"</literal> to let an admin user specify a
scheduler hint to direct a build to a particular cell.</para>
</section>
<section xml:id="cell-config-optional-json">
<title>Optional cell configuration</title>
<para>Cells currently keeps all inter-cell communication data, including
user names and passwords, in the database. This is undesirable and
unnecessary since cells data isn't updated very frequently. Instead,
create a JSON file to input cells data specified via a
<code>[cells]cells_config</code> option. When specified, the
database is no longer consulted when reloading cells data. The file
will need the columns present in the Cell model (excluding common
database fields and the <code>id</code> column). The queue
connection information must be specified through a
<code>transport_url</code> field, instead of
<code>username</code>, <code>password</code>, and so on. The
<code>transport_url</code> has the following form:</para>
<programlisting>
rabbit://&lt;username>:&lt;password>@&lt;hostname>:&lt;port>/&lt;virtual_host>
</programlisting>
<para>The scheme can be either <literal>qpid</literal> or <literal>rabbit</literal>, as shown previously. The following sample shows this optional configuration:</para>
<programlisting language="json"><xi:include href="samples/cells-resp.json" parse="text"/></programlisting>
</section>
</section>