cloudstack/docs/en-US/autoscale.xml

<?xml version='1.0' encoding='utf-8' ?>
<!DOCTYPE section PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN" "http://www.oasis-open.org/docbook/xml/4.5/docbookx.dtd" [
<!ENTITY % BOOK_ENTITIES SYSTEM "cloudstack.ent">
%BOOK_ENTITIES;
]>

<!-- Licensed to the Apache Software Foundation (ASF) under one
    or more contributor license agreements.  See the NOTICE file
    distributed with this work for additional information
    regarding copyright ownership.  The ASF licenses this file
    to you under the Apache License, Version 2.0 (the
    "License"); you may not use this file except in compliance
    with the License.  You may obtain a copy of the License at
    http://www.apache.org/licenses/LICENSE-2.0
    Unless required by applicable law or agreed to in writing,
    software distributed under the License is distributed on an
    "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
    KIND, either express or implied.  See the License for the
    specific language governing permissions and limitations
    under the License.
-->
<section id="autoscale">
  <title>Configuring AutoScale</title>
  <para>AutoScaling allows you to scale your back-end services or application VMs up or down
    seamlessly and automatically according to the conditions you define. With AutoScaling enabled,
    you can ensure that the number of VMs you are using seamlessly scale up when demand increases,
    and automatically decreases when demand subsides. Thus it helps you save compute costs by
    terminating underused VMs automatically and launching new VMs when you need them, without the
    need for manual intervention.</para>
  <para>NetScaler AutoScaling is designed to seamlessly launch or terminate VMs based on
    user-defined conditions. Conditions for triggering a scaleup or scaledown action can vary from a
    simple use case like monitoring the CPU usage of a server to a complex use case of monitoring a
    combination of server's responsiveness and its CPU usage. For example, you can configure
    AutoScaling to launch an additional VM whenever CPU usage exceeds 80 percent for 15 minutes, or
    to remove a VM whenever CPU usage is less than 20 percent for 30 minutes.</para>
  <para>&PRODUCT; uses the NetScaler load balancer to monitor all aspects of a system's health and
    work in unison with &PRODUCT; to initiate scale-up or scale-down actions.</para>
  <note>
    <para>AutoScale is supported on NetScaler Release 10 Build 73.e and beyond.</para>
  </note>
  <formalpara>
    <title>Prerequisites</title>
    <para>Before you configure an AutoScale rule, consider the following:</para>
  </formalpara>
  <itemizedlist>
    <listitem>
      <para>Ensure that the necessary template is prepared before configuring AutoScale. When a VM
        is deployed by using a template and when it comes up, the application should be up and
        running.</para>
      <note>
        <para>If the application is not running, the NetScaler device considers the VM as
          ineffective and continues provisioning the VMs unconditionally until the resource limit is
          exhausted.</para>
      </note>
    </listitem>
    <listitem>
      <para>Deploy the templates you prepared. Ensure that the applications come up on the first
        boot and is ready to take the traffic. Observe the time requires to deploy the template.
        Consider this time when you specify the quiet time while configuring AutoScale.</para>
    </listitem>
    <listitem>
      <para>The AutoScale feature supports the SNMP counters that can be used to define conditions
        for taking scale up or scale down actions. To monitor the SNMP-based counter, ensure that
        the SNMP agent is installed in the template used for creating the AutoScale VMs, and the
        SNMP operations work with the configured SNMP community and port by using standard SNMP
        managers. For example, see <xref linkend="configure-snmp-rhel"/> to configure SNMP on a RHEL
        machine.</para>
    </listitem>
    <listitem>
      <para>Ensure that the endpointe.url parameter present in the Global Settings is set to the
        Management Server API URL. For example, http://10.102.102.22:8080/client/api. In a
        multi-node Management Server deployment, use the virtual IP address configured in the load
        balancer for the management server’s cluster. Additionally, ensure that the NetScaler device
        has access to this IP address to provide AutoScale support.</para>
      <para>If you update the endpointe.url, disable the AutoScale functionality of the load
        balancer rules in the system, then enable them back to reflect the changes. For more
        information see <xref linkend="update-autoscale"/></para>
    </listitem>
    <listitem>
      <para>If the API Key and Secret Key are regenerated for an AutoScale user, ensure that the
        AutoScale functionality of the load balancers that the user participates in are disabled and
        then enabled to reflect the configuration changes in the NetScaler.</para>
    </listitem>
    <listitem>
      <para>In an advanced Zone, ensure that at least one VM should be present before configuring a
        load balancer rule with AutoScale. Having one VM in the network ensures that the network is
        in implemented state for configuring AutoScale.</para>
    </listitem>
  </itemizedlist>
  <formalpara>
    <title>Configuration</title>
    <para>Specify the following:</para>
  </formalpara>
  <mediaobject>
    <imageobject>
      <imagedata fileref="./images/autoscale-config.png"/>
    </imageobject>
    <textobject>
      <phrase>autoscaleateconfig.png: Configuring AutoScale</phrase>
    </textobject>
  </mediaobject>
  <itemizedlist>
    <listitem>
      <para><emphasis role="bold">Template</emphasis>: A template consists of a base OS image and
        application. A template is used to provision the new instance of an application on a scaleup
        action. When a VM is deployed from a template, the VM can start taking the traffic from the
        load balancer without any admin intervention. For example, if the VM is deployed for a Web
        service, it should have the Web server running, the database connected, and so on.</para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Compute offering</emphasis>: A predefined set of virtual hardware
        attributes, including CPU speed, number of CPUs, and RAM size, that the user can select when
        creating a new virtual machine instance. Choose one of the compute offerings to be used
        while provisioning a VM instance as part of scaleup action.</para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Min Instance</emphasis>: The minimum number of active VM instances
        that is assigned to a load balancing rule. The active VM instances are the application
        instances that are up and serving the traffic, and are being load balanced. This parameter
        ensures that a load balancing rule has at least the configured number of active VM instances
        are available to serve the traffic.</para>
      <note>
        <para>If an application, such as SAP, running on a VM instance is down for some reason, the
          VM is then not counted as part of Min Instance parameter, and the AutoScale feature
          initiates a scaleup action if the number of active VM instances is below the configured
          value. Similarly, when an application instance comes up from its earlier down state, this
          application instance is counted as part of the active instance count and the AutoScale
          process initiates a scaledown action when the active instance count breaches the Max
          instance value.</para>
      </note>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Max Instance</emphasis>: Maximum number of active VM instances
        that <emphasis role="bold">should be assigned to </emphasis>a load balancing rule. This
        parameter defines the upper limit of active VM instances that can be assigned to a load
        balancing rule.</para>
      <para>Specifying a large value for the maximum instance parameter might result in provisioning
        large number of VM instances, which in turn leads to a single load balancing rule exhausting
        the VM instances limit specified at the account or domain level.</para>
      <note>
        <para>If an application, such as SAP, running on a VM instance is down for some reason, the
          VM is not counted as part of Max Instance parameter. So there may be scenarios where the
          number of VMs provisioned for a scaleup action might be more than the configured Max
          Instance value. Once the application instances in the VMs are up from an earlier down
          state, the AutoScale feature starts aligning to the configured Max Instance value.</para>
      </note>
    </listitem>
  </itemizedlist>
  <para>Specify the following scale-up and scale-down policies:</para>
  <itemizedlist>
    <listitem>
      <para><emphasis role="bold">Duration</emphasis>: The duration, in seconds, for which the
        conditions you specify must be true to trigger a scaleup action. The conditions defined
        should hold true for the entire duration you specify for an AutoScale action to be invoked.
      </para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Counter</emphasis>: The performance counters expose the state of
        the monitored instances. By default, &PRODUCT; offers four performance counters: Three SNMP
        counters and one NetScaler counter. The SNMP counters are Linux User CPU, Linux System CPU,
        and Linux CPU Idle. The NetScaler counter is ResponseTime. The root administrator can add
        additional counters into &PRODUCT; by using the &PRODUCT; API. </para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Operator</emphasis>: The following five relational operators are
        supported in AutoScale feature: Greater than, Less than, Less than or equal to, Greater than
        or equal to, and Equal to.</para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Threshold</emphasis>: Threshold value to be used for the counter.
        Once the counter defined above breaches the threshold value, the AutoScale feature initiates
        a scaleup or scaledown action.</para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Add</emphasis>: Click Add to add the condition.</para>
    </listitem>
  </itemizedlist>
  <para>Additionally, if you want to configure the advanced settings, click Show advanced settings,
    and specify the following:</para>
  <itemizedlist>
    <listitem>
      <para><emphasis role="bold">Polling interval</emphasis>: Frequency in which the conditions,
        combination of counter, operator and threshold, are to be evaluated before taking a scale up
        or down action. The default polling interval is 30 seconds.</para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Quiet Time</emphasis>: This is the cool down period after an
        AutoScale action is initiated. The time includes the time taken to complete provisioning a
        VM instance from its template and the time taken by an application to be ready to serve
        traffic. This quiet time allows the fleet to come up to a stable state before any action can
        take place. The default is 300 seconds.</para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Destroy VM Grace Period</emphasis>: The duration in seconds, after
        a scaledown action is initiated, to wait before the VM is destroyed as part of scaledown
        action. This is to ensure graceful close of any pending sessions or transactions being
        served by the VM marked for destroy. The default is 120 seconds.</para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Security Groups</emphasis>: Security groups provide a way to
        isolate traffic to the VM instances. A security group is a group of VMs that filter their
        incoming and outgoing traffic according to a set of rules, called ingress and egress rules.
        These rules filter network traffic according to the IP address that is attempting to
        communicate with the VM.</para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Disk Offerings</emphasis>: A predefined set of disk size for
        primary data storage. </para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">SNMP Community</emphasis>: The SNMP community string to be used by
        the NetScaler device to query the configured counter value from the provisioned VM
        instances. Default is public.</para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">SNMP Port</emphasis>: The port number on which the SNMP agent that
        run on the provisioned VMs is listening. Default port is 161. </para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">User</emphasis>: This is the user that the NetScaler device use to
        invoke scaleup and scaledown API calls to the cloud. If no option is specified, the user who
        configures AutoScaling is applied. Specify another user name to override.</para>
    </listitem>
    <listitem>
      <para><emphasis role="bold">Apply</emphasis>: Click Apply to create the AutoScale
        configuration.</para>
    </listitem>
  </itemizedlist>
  <formalpara>
    <title>Disabling and Enabling an AutoScale Configuration</title>
    <para>If you want to perform any maintenance operation on the AutoScale VM instances, disable
      the AutoScale configuration. When the AutoScale configuration is disabled, no scaleup or
      scaledown action is performed. You can use this downtime for the maintenance activities. To
      disable the AutoScale configuration, click the Disable AutoScale<inlinemediaobject>
        <imageobject>
          <imagedata fileref="./images/enable-disable-autoscale.png"/>
        </imageobject>
        <textobject>
          <phrase>EnableDisable.png: button to enable or disable AutoScale.</phrase>
        </textobject>
      </inlinemediaobject>button.</para>
  </formalpara>
  <para>The button toggles between enable and disable, depending on whether AutoScale is currently
    enabled or not. After the maintenance operations are done, you can enable the AutoScale
    configuration back. To enable, open the AutoScale configuration page again, then click the
    Enable AutoScale<inlinemediaobject>
      <imageobject>
        <imagedata fileref="./images/enable-disable-autoscale.png"/>
      </imageobject>
      <textobject>
        <phrase>EnableDisable.png: button to enable or disable AutoScale.</phrase>
      </textobject>
    </inlinemediaobject>button.</para>
  <formalpara id="update-autoscale">
    <title>Updating an AutoScale Configuration</title>
    <para>You can update the various parameters and add or delete the conditions in a scaleup or
      scaledown rule. Before you update an AutoScale configuration, ensure that you disable the
      AutoScale load balancer rule by clicking the Disable AutoScale button.</para>
  </formalpara>
  <para>After you modify the required AutoScale parameters, click Apply. To apply the new AutoScale
    policies, open the AutoScale configuration page again, then click the Enable AutoScale
    button.</para>
  <formalpara>
    <title>Runtime Considerations</title>
    <para/>
  </formalpara>
  <itemizedlist>
    <listitem>
      <para>An administrator should not assign a VM to a load balancing rule which is configured for
        AutoScale.</para>
    </listitem>
    <listitem>
      <para>Before a VM provisioning is completed if NetScaler is shutdown or restarted, the
        provisioned VM cannot be a part of the load balancing rule though the intent was to assign
        it to a load balancing rule. To workaround, rename the AutoScale provisioned VMs based on
        the rule name or ID so at any point of time the VMs can be reconciled to its load balancing
        rule.</para>
    </listitem>
    <listitem>
      <para>Making API calls outside the context of AutoScale, such as destroyVM, on an autoscaled
        VM leaves the load balancing configuration in an inconsistent state. Though VM is destroyed
        from the load balancer rule, NetScaler continues to show the VM as a service assigned to a
        rule.</para>
    </listitem>
  </itemizedlist>
</section>