impala/docs/topics/impala_proxy.xml

<?xml version="1.0" encoding="UTF-8"?>
<!--
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.
-->
<!DOCTYPE concept PUBLIC "-//OASIS//DTD DITA Concept//EN" "concept.dtd">
<concept id="proxy">

  <title>Using Impala through a Proxy for High Availability</title>

  <titlealts audience="PDF">

    <navtitle>Load-Balancing Proxy for HA</navtitle>

  </titlealts>

  <prolog>
    <metadata>
      <data name="Category" value="High Availability"/>
      <data name="Category" value="Impala"/>
      <data name="Category" value="Network"/>
      <data name="Category" value="Proxy"/>
      <data name="Category" value="Administrators"/>
      <data name="Category" value="Developers"/>
      <data name="Category" value="Data Analysts"/>
    </metadata>
  </prolog>

  <conbody>

    <p>
      For most clusters that have multiple users and production availability requirements, you
      might want to set up a load-balancing proxy server to relay requests to and from Impala.
    </p>

    <p>
      Set up a software package of your choice to perform these functions.
    </p>

    <note>
      <p conref="../shared/impala_common.xml#common/statestored_catalogd_ha_blurb"/>
    </note>

    <p outputclass="toc inpage"/>

  </conbody>

  <concept id="proxy_overview">

    <title>Overview of Proxy Usage and Load Balancing for Impala</title>

    <prolog>
      <metadata>
        <data name="Category" value="Concepts"/>
      </metadata>
    </prolog>

    <conbody>

      <p>
        Using a load-balancing proxy server for Impala has the following advantages:
      </p>

      <ul>
        <li>
          Applications connect to a single well-known host and port, rather than keeping track
          of the hosts where the <cmdname>impalad</cmdname> daemon is running.
        </li>

        <li>
          If any host running the <cmdname>impalad</cmdname> daemon becomes unavailable,
          application connection requests still succeed because you always connect to the proxy
          server rather than a specific host running the <cmdname>impalad</cmdname> daemon.
        </li>

        <li> The coordinator node for each Impala query potentially requires
          more memory and CPU cycles than the other nodes that process the
          query. The proxy server can issue queries so that each connection uses
          a different coordinator node. This load-balancing technique lets the
            <cmdname>impalad</cmdname> nodes share this additional work, rather
          than concentrating it on a single machine. </li>
      </ul>

      <p>
        The following setup steps are a general outline that apply to any load-balancing proxy
        software:
      </p>

      <ol>
        <li>
          Select and download the load-balancing proxy software or other load-balancing hardware
          appliance. It should only need to be installed and configured on a single host,
          typically on an edge node.
        </li>

        <li>
          Configure the load balancer (typically by editing a configuration file). In
          particular:
          <ul>
            <li>
              To relay Impala requests back and forth, set up a port that the load balancer will
              listen on.
            </li>

            <li>
              Select a load balancing algorithm. See
              <xref
                href="#proxy_balancing" format="dita"/> for load balancing
              algorithm options.
            </li>

            <li>
              For Kerberized clusters, follow the instructions in
              <xref
                href="impala_proxy.xml#proxy_kerberos"/>.
            </li>
          </ul>
        </li>

        <li>
          If you are using Hue or JDBC-based applications, you typically set up load balancing
          for both ports 21000 and 21050 because these client applications connect through port
          21050 while the <cmdname>impala-shell</cmdname> command connects through port 21000.
          See <xref href="impala_ports.xml#ports"/> for when to use port 21000, 21050, or
          another value depending on what type of connections you are load balancing.
        </li>

        <li>
          Run the load-balancing proxy server, pointing it at the configuration file that you
          set up.
        </li>

        <li>
          For any scripts, jobs, or configuration settings for applications that formerly
          connected to a specific <cmdname>impalad</cmdname> to run Impala SQL statements,
          change the connection information (such as the <codeph>-i</codeph> option in
          <cmdname>impala-shell</cmdname>) to point to the load balancer instead.
        </li>
      </ol>

      <note>
        The following sections use the HAProxy software as a representative example of a load
        balancer that you can use with Impala.
      </note>

    </conbody>

  </concept>

  <concept id="proxy_balancing" rev="">

    <title>Choosing the Load-Balancing Algorithm</title>

    <conbody>

      <p>
        Load-balancing software offers a number of algorithms to distribute requests. Each
        algorithm has its own characteristics that make it suitable in some situations but not
        others.
      </p>

      <dl>
        <dlentry>

          <dt>
            Leastconn
          </dt>

          <dd>
            Connects sessions to the coordinator with the fewest connections, to balance the
            load evenly. Typically used for workloads consisting of many independent,
            short-running queries. In configurations with only a few client machines, this
            setting can avoid having all requests go to only a small set of coordinators.
          </dd>

          <dd>
            Recommended for Impala with F5.
          </dd>

        </dlentry>

        <dlentry>

          <dt>
            Source IP Persistence
          </dt>

          <dd>
            <p>
              Sessions from the same IP address always go to the same coordinator. A good choice
              for Impala workloads containing a mix of queries and DDL statements, such as
              <codeph>CREATE TABLE</codeph> and <codeph>ALTER TABLE</codeph>. Because the
              metadata changes from a DDL statement take time to propagate across the cluster,
              prefer to use the Source IP Persistence in this case. If you are unable to choose
              Source IP Persistence, run the DDL and subsequent queries that depend on the
              results of the DDL through the same session, for example by running
              <codeph>impala-shell -f <varname>script_file</varname></codeph> to submit several
              statements through a single session.
            </p>
          </dd>

          <dd>
            <p>
              Required for setting up high availability with Hue.
            </p>
          </dd>

        </dlentry>

        <dlentry>

          <dt>
            Round-robin
          </dt>

          <dd>
            Distributes connections to all coordinator nodes. Typically not recommended for
            Impala.
          </dd>

        </dlentry>
      </dl>

      <p>
        You might need to perform benchmarks and load testing to determine which setting is
        optimal for your use case. Always set up using two load-balancing algorithms: Source IP
        Persistence for Hue and Leastconn for others.
      </p>

    </conbody>

  </concept>

  <concept id="proxy_kerberos">

    <title>Special Proxy Considerations for Clusters Using Kerberos</title>

    <prolog>
      <metadata>
        <data name="Category" value="Security"/>
        <data name="Category" value="Kerberos"/>
        <data name="Category" value="Authentication"/>
        <data name="Category" value="Proxy"/>
      </metadata>
    </prolog>

    <conbody>

      <p>
        In a cluster using Kerberos, applications check host credentials to verify that the host
        they are connecting to is the same one that is actually processing the request.
      </p>

      <p>
        In <keyword keyref="impala211_full">Impala 2.11</keyword> and lower versions, once you
        enable a proxy server in a Kerberized cluster, users will not be able to connect to
        individual impala daemons directly from impala-shell.
      </p>

      <p>
        In <keyword keyref="impala212_full">Impala 2.12</keyword> and higher versions, when you
        enable a proxy server in a Kerberized cluster, users have an option to connect to Impala
        daemons directly from <cmdname>impala-shell</cmdname> using the <codeph>-b</codeph> /
        <codeph>--kerberos_host_fqdn</codeph> <cmdname>impala-shell</cmdname> flag. This option
        can be used for testing or troubleshooting purposes, but not recommended for live
        production environments as it defeats the purpose of a load balancer/proxy.
      </p>

      <p>
        Example:
<codeblock>
impala-shell -i impalad-1.mydomain.com -k -b loadbalancer-1.mydomain.com
</codeblock>
      </p>

      <p>
        Alternatively, with the fully qualified configurations:
<codeblock>impala-shell --impalad=impalad-1.mydomain.com:21000 --kerberos --kerberos_host_fqdn=loadbalancer-1.mydomain.com</codeblock>
      </p>

      <p>
        See <xref href="impala_shell_options.xml#shell_options"/> for information about the
        option.
      </p>

      <p>
        To validate the load-balancing proxy server, perform these extra Kerberos setup steps:
      </p>

      <ol>
        <li>
          This section assumes you are starting with a Kerberos-enabled cluster. See
          <xref href="impala_kerberos.xml#kerberos"/> for instructions for setting up Impala
          with Kerberos. See <xref keyref="cdh_sg_kerberos_prin_keytab_deploy"/> for general
          steps to set up Kerberos.
        </li>

        <li>
          Choose the host you will use for the proxy server. Based on the Kerberos setup
          procedure, it should already have an entry
          <codeph>impala/<varname>proxy_host</varname>@<varname>realm</varname></codeph> in its
          <filepath>keytab</filepath>. If not, go back over the initial Kerberos configuration
          steps for the <filepath>keytab</filepath> on each host running the
          <cmdname>impalad</cmdname> daemon.
        </li>

        <li>
          Copy the <filepath>keytab</filepath> file from the proxy host to all other hosts in
          the cluster that run the <cmdname>impalad</cmdname> daemon. Put the
          <filepath>keytab</filepath> file in a secure location on each of these other hosts.
        </li>

        <li>
          Add an entry
          <codeph>impala/<varname>actual_hostname</varname>@<varname>realm</varname></codeph> to
          the <filepath>keytab</filepath> on each host running the <cmdname>impalad</cmdname>
          daemon.
        </li>

        <li>
          For each <cmdname>impalad</cmdname> node, merge the existing
          <filepath>keytab</filepath> with the proxy’s <filepath>keytab</filepath> using
          <cmdname>ktutil</cmdname>, producing a new <filepath>keytab</filepath> file. For
          example:
<codeblock>$ ktutil
  ktutil: read_kt proxy.keytab
  ktutil: read_kt impala.keytab
  ktutil: write_kt proxy_impala.keytab
  ktutil: quit</codeblock>
        </li>

        <li>
          To verify that the <filepath>keytabs</filepath> are merged, run the command:
<codeblock>
klist -k <varname>keytabfile</varname>
</codeblock>
          The command lists the credentials for both <codeph>principal</codeph> and
          <codeph>be_principal</codeph> on all nodes.
        </li>

        <li>
          Make sure that the <codeph>impala</codeph> user has the permission to read this merged
          <filepath>keytab</filepath> file.
        </li>

        <li>
          For each coordinator <codeph>impalad</codeph> host in the cluster that participates in
          the load balancing, add the following configuration options to receive client
          connections coming through the load balancer proxy server:
<codeblock>
--principal=impala/<varname>proxy_host@realm</varname>
  --be_principal=impala/<varname>actual_host@realm</varname>
  --keytab_file=<varname>path_to_merged_keytab</varname>
</codeblock>
          <p>
            The <codeph>--principal</codeph> setting prevents a client from connecting to a
            coordinator <codeph>impalad</codeph> using a principal other than one specified.
          </p>

          <note>
            Every host has different <codeph>--be_principal</codeph> because the actual host
            name is different on each host. Specify the fully qualified domain name (FQDN) for
            the proxy host, not the IP address. Use the exact FQDN as returned by a reverse DNS
            lookup for the associated IP address.
          </note>
        </li>

        <li>
          Restart Impala to make the changes take effect. Restart the <cmdname>impalad</cmdname>
          daemons on all hosts in the cluster, as well as the <cmdname>statestored</cmdname> and
          <cmdname>catalogd</cmdname> daemons.
        </li>
      </ol>

      <section id="section_fjz_mfn_yjb">

        <title>Client Connection to Proxy Server in Kerberized Clusters</title>

        <p>
          When a client connect to Impala, the service principal specified by the client must
          match the <codeph>-principal</codeph> setting of the Impala proxy server. And the
          client should connect to the proxy server port.
        </p>

        <p>
          In <filepath>hue.ini</filepath>, set the following to configure Hue to automatically
          connect to the proxy server:
        </p>

<codeblock>[impala]
server_host=<varname>proxy_host</varname>
impala_principal=impala/<varname>proxy_host</varname></codeblock>

        <p>
          The following are the JDBC connection string formats when connecting through the load
          balancer with the load balancer's host name in the principal:
        </p>

<codeblock>jdbc:hive2://<varname>proxy_host</varname>:<varname>load_balancer_port</varname>/;principal=impala/_HOST@<varname>realm</varname>
jdbc:hive2://<varname>proxy_host</varname>:<varname>load_balancer_port</varname>/;principal=impala/<varname>proxy_host</varname>@<varname>realm</varname></codeblock>

        <p>
          When starting <cmdname>impala-shell</cmdname>, specify the service principal via the
          <codeph>-b</codeph> or <codeph>--kerberos_host_fqdn</codeph> flag.
        </p>

      </section>

    </conbody>

  </concept>

  <concept id="proxy_tls">

    <title>Special Proxy Considerations for TLS/SSL Enabled Clusters</title>

    <prolog>
      <metadata>
        <data name="Category" value="Security"/>
        <data name="Category" value="TLS"/>
        <data name="Category" value="Authentication"/>
        <data name="Category" value="Proxy"/>
      </metadata>
    </prolog>

    <conbody>

      <p>
        When TLS/SSL is enabled for Impala, the client application, whether impala-shell, Hue,
        or something else, expects the certificate common name (CN) to match the hostname that
        it is connected to. With no load balancing proxy server, the hostname and certificate CN
        are both that of the <codeph>impalad</codeph> instance. However, with a proxy server,
        the certificate presented by the <codeph>impalad</codeph> instance does not match the
        load balancing proxy server hostname. If you try to load-balance a TLS/SSL-enabled
        Impala installation without additional configuration, you see a certificate mismatch
        error when a client attempts to connect to the load balancing proxy host.
      </p>

      <p>
        You can configure a proxy server in several ways to load balance TLS/SSL enabled Impala:
      </p>

      <dl>
        <dlentry>
          <dt> TLS/SSL Bridging</dt>
          <dd> In this configuration, the proxy server presents a TLS/SSL
            certificate to the client, decrypts the client request, then
            re-encrypts the request before sending it to the backend
              <codeph>impalad</codeph>. The client and server certificates can
            be managed separately. The request or resulting payload is encrypted
            in transit at all times. </dd>

        </dlentry>

        <dlentry>

          <dt>
            TLS/SSL Passthrough
          </dt>

          <dd>
            In this configuration, traffic passes through to the backend
            <codeph>impalad</codeph> instance with no interaction from the load balancing proxy
            server. Traffic is still encrypted end-to-end.
          </dd>

          <dd>
            The same server certificate, utilizing either wildcard or Subject Alternate Name
            (SAN), must be installed on each <codeph>impalad</codeph> instance.
          </dd>

        </dlentry>

        <dlentry>

          <dt>
            TLS/SSL Offload
          </dt>

          <dd>
            In this configuration, all traffic is decrypted on the load balancing proxy server,
            and traffic between the backend <codeph>impalad</codeph> instances is unencrypted.
            This configuration presumes that cluster hosts reside on a trusted network and only
            external client-facing communication need to be encrypted in-transit.
          </dd>

        </dlentry>
      </dl>

      <p>
        Refer to your load balancer documentation for the steps to set up Impala and the load
        balancer using one of the options above.
      </p>

    </conbody>

  </concept>

  <concept id="tut_proxy">

    <title>Example of Configuring HAProxy Load Balancer for Impala</title>

    <prolog>
      <metadata>
        <data name="Category" value="Configuring"/>
      </metadata>
    </prolog>

    <conbody>

      <p>
        If you are not already using a load-balancing proxy, you can experiment with
        <xref href="http://haproxy.1wt.eu/" scope="external" format="html">HAProxy</xref> a
        free, open source load balancer. This example shows how you might install and configure
        that load balancer on a Red Hat Enterprise Linux system.
      </p>

      <ul>
        <li>
          <p>
            Install the load balancer:
          </p>
<codeblock>yum install haproxy</codeblock>
        </li>

        <li>
          <p>
            Set up the configuration file: <filepath>/etc/haproxy/haproxy.cfg</filepath>. See
            the following section for a sample configuration file.
          </p>
        </li>

        <li>
          <p>
            Run the load balancer (on a single host, preferably one not running
            <cmdname>impalad</cmdname>):
          </p>
<codeblock>/usr/sbin/haproxy –f /etc/haproxy/haproxy.cfg</codeblock>
        </li>

        <li>
          <p>
            In <cmdname>impala-shell</cmdname>, JDBC applications, or ODBC applications, connect
            to the listener port of the proxy host, rather than port 21000 or 21050 on a host
            actually running <cmdname>impalad</cmdname>. The sample configuration file sets
            haproxy to listen on port 25003, therefore you would send all requests to
            <codeph><varname>haproxy_host</varname>:25003</codeph>.
          </p>
        </li>
      </ul>

      <p>
        This is the sample <filepath>haproxy.cfg</filepath> used in this example:
      </p>

<codeblock>global
    # To have these messages end up in /var/log/haproxy.log you will
    # need to:
    #
    # 1) configure syslog to accept network log events.  This is done
    #    by adding the '-r' option to the SYSLOGD_OPTIONS in
    #    /etc/sysconfig/syslog
    #
    # 2) configure local2 events to go to the /var/log/haproxy.log
    #   file. A line like the following can be added to
    #   /etc/sysconfig/syslog
    #
    #    local2.*                       /var/log/haproxy.log
    #
    log         127.0.0.1 local0
    log         127.0.0.1 local1 notice
    chroot      /var/lib/haproxy
    pidfile     /var/run/haproxy.pid
    maxconn     4000
    user        haproxy
    group       haproxy
    daemon

    # turn on stats unix socket
    #stats socket /var/lib/haproxy/stats

#---------------------------------------------------------------------
# common defaults that all the 'listen' and 'backend' sections will
# use if not designated in their block
#
# You might need to adjust timing values to prevent timeouts.
#
# The timeout values should be dependant on how you use the cluster
# and how long your queries run.
#---------------------------------------------------------------------
defaults
    mode                    http
    log                     global
    option                  httplog
    option                  dontlognull
    option http-server-close
    option forwardfor       except 127.0.0.0/8
    option                  redispatch
    retries                 3
    maxconn                 3000
    timeout connect 5000
    timeout client 3600s
    timeout server 3600s

#
# This sets up the admin page for HA Proxy at port 25002.
#
listen stats :25002
    balance
    mode http
    stats enable
    stats auth <varname>username</varname>:<varname>password</varname>

# Setup for Impala.
# Impala client connect to load_balancer_host:25003.
# HAProxy will balance connections among the list of servers listed below.
# The list of Impalad is listening at port 21000 for beeswax (impala-shell) or original ODBC driver.
# For JDBC or ODBC version 2.x driver, use port 21050 instead of 21000.
listen impala :25003
    mode tcp
    option tcplog
    balance leastconn

    server <varname>symbolic_name_1</varname> impala-host-1.example.com:21000 check
    server <varname>symbolic_name_2</varname> impala-host-2.example.com:21000 check
    server <varname>symbolic_name_3</varname> impala-host-3.example.com:21000 check
    server <varname>symbolic_name_4</varname> impala-host-4.example.com:21000 check

# Setup for Hue or other JDBC-enabled applications.
# In particular, Hue requires sticky sessions.
# The application connects to load_balancer_host:21051, and HAProxy balances
# connections to the associated hosts, where Impala listens for
# JDBC requests at port 21050.
listen impalajdbc :21051
    mode tcp
    option tcplog
    balance source

    server <varname>symbolic_name_5</varname> impala-host-1.example.com:21050 check
    server <varname>symbolic_name_6</varname> impala-host-2.example.com:21050 check
    server <varname>symbolic_name_7</varname> impala-host-3.example.com:21050 check
    server <varname>symbolic_name_8</varname> impala-host-4.example.com:21050 check
</codeblock>

      <note type="important">
        Hue requires the <codeph>check</codeph> option at end of each line in the above file to
        ensure HAProxy can detect any unreachable <cmdname>Impalad</cmdname> server, and
        failover can be successful. Without the TCP check, you may hit an error when the
        <cmdname>impalad</cmdname> daemon to which Hue tries to connect is down.
      </note>

      <note conref="../shared/impala_common.xml#common/proxy_jdbc_caveat"/>

    </conbody>

  </concept>

</concept>