forked from docs/doc-exports
Yang, Tong
6182f91ba8
Reviewed-by: Hasko, Vladimir <vladimir.hasko@t-systems.com> Co-authored-by: Yang, Tong <yangtong2@huawei.com> Co-committed-by: Yang, Tong <yangtong2@huawei.com>
25 lines
5.4 KiB
HTML
25 lines
5.4 KiB
HTML
<a name="mrs_01_1681"></a><a name="mrs_01_1681"></a>
|
|
|
|
<h1 class="topictitle1">Configuring the Observer NameNode to Process Read Requests</h1>
|
|
<div id="body1595904096782"><div class="section" id="mrs_01_1681__se9608011680e423ca403d5207c374daa"><h4 class="sectiontitle">Scenario</h4><p id="mrs_01_1681__p17401421813">In an HDFS cluster configured with HA, the active NameNode processes all client requests, and the standby NameNode reserves the latest metadata and block location information. However, in this architecture, the active NameNode is the bottleneck of client request processing. This bottleneck is more obvious in clusters with a large number of requests.</p>
|
|
<p id="mrs_01_1681__p4733591402">To address this issue, a new NameNode is introduced: an observer NameNode. Similar to the standby NameNode, the observer NameNode also reserves the latest metadata information and block location information. In addition, the observer NameNode can process read requests from clients in the same way as the active NameNode. In typical HDFS clusters with many read requests, the observer NameNode can be used to process read requests, reducing the active NameNode load and improving the cluster capability of processing requests.</p>
|
|
<div class="note" id="mrs_01_1681__note17408121619136"><img src="public_sys-resources/note_3.0-en-us.png"><span class="notetitle"> </span><div class="notebody"><p id="mrs_01_1681__p788202018132">This section applies to MRS 3.<em id="mrs_01_1681__i5543022405157">x</em> or later.</p>
|
|
</div></div>
|
|
</div>
|
|
<div class="section" id="mrs_01_1681__s96b8a9dc800a44b7b35e3a2d8fafa0e1"><h4 class="sectiontitle">Impact on the System</h4><ul id="mrs_01_1681__ul10686111983619"><li id="mrs_01_1681__li126872191367">The active NameNode load can be reduced and the capability of HDFS cluster processing requests can be improved, which is especially obvious for large clusters.</li><li id="mrs_01_1681__li278002117364">The client application configuration needs to be updated.</li></ul>
|
|
</div>
|
|
<div class="section" id="mrs_01_1681__sa8a135a114cf4cbc8674242bd0cfebd7"><h4 class="sectiontitle">Prerequisites</h4><ul id="mrs_01_1681__ul6387132716574"><li id="mrs_01_1681__la8a352e20c414015a0e60f8023ecc0a5">The HDFS cluster has been installed, the active and standby NameNodes are running properly, and the HDFS service is normal.</li><li id="mrs_01_1681__li196291144105015">The <strong id="mrs_01_1681__b6780622085157">${BIGDATA_DATA_HOME}/namenode</strong> partition has been created on the node where the observer NameNode is to be installed.</li></ul>
|
|
</div>
|
|
<div class="section" id="mrs_01_1681__section11217152712020"><h4 class="sectiontitle">Procedure</h4><p id="mrs_01_1681__p116078478421">The following steps describe how to configure the observer NameNode of a hacluster and enable it to process read requests. If there are multiple pairs of NameServices in the cluster and they are all in use, perform the following steps to configure the observer NameNode for each pair.</p>
|
|
<ol id="mrs_01_1681__ol14972170134210"><li id="mrs_01_1681__li20972170164215"><span>Log in to FusionInsight Manager.</span></li><li id="mrs_01_1681__li125071111539"><span>Choose <span id="mrs_01_1681__text1751171119531"><strong id="mrs_01_1681__b1328853205157">Cluster</strong> > </span><em id="mrs_01_1681__i6137876195157">Name of the desired cluster</em> > <strong id="mrs_01_1681__b3794114635157">Services</strong> > <strong id="mrs_01_1681__b6767286455157">HDFS</strong> > <strong id="mrs_01_1681__b19250096445157">NameService Management</strong>.</span></li><li id="mrs_01_1681__li5929133819477"><span>Click <strong id="mrs_01_1681__b19039094795157">Add</strong> next to <strong id="mrs_01_1681__b3137404115157">hacluster</strong>.</span></li><li id="mrs_01_1681__li169891275112"><span>On the <strong id="mrs_01_1681__b9696140645157">Add NameNode</strong> page, set <strong id="mrs_01_1681__b7075855745157">NameNode type</strong> to <strong id="mrs_01_1681__b13543377485157">Observer</strong> and click <strong id="mrs_01_1681__b1573278665157">Next</strong>.</span></li><li id="mrs_01_1681__li141141329549"><span>On the <strong id="mrs_01_1681__b12341237835157">Assign Role</strong> page, select the planned host, add the observer NameNode, and click <strong id="mrs_01_1681__b19057901705157">Next</strong>.</span><p><div class="note" id="mrs_01_1681__note1662797113818"><img src="public_sys-resources/note_3.0-en-us.png"><span class="notetitle"> </span><div class="notebody"><p id="mrs_01_1681__p10628177193811">A maximum of five observer NameNodes can be added to each pair of NameServices.</p>
|
|
</div></div>
|
|
</p></li><li id="mrs_01_1681__li6102155713"><span>On the configuration page, configure the storage directory and port number of the NameNode as planned and click <strong id="mrs_01_1681__b3821924745157">Next</strong>.</span></li><li id="mrs_01_1681__li107303481708"><span>Confirm the information, click <strong id="mrs_01_1681__b8316340455157">Submit</strong>, and wait until the installation of the observer NameNode is complete.</span></li></ol><ol start="8" id="mrs_01_1681__ol6775137111713"><li id="mrs_01_1681__li2775337161719"><span>Restart the upper-layer components that depend on HDFS, update the client application configuration, and restart the client application.</span></li></ol>
|
|
</div>
|
|
</div>
|
|
<div>
|
|
<div class="familylinks">
|
|
<div class="parentlink"><strong>Parent topic:</strong> <a href="mrs_01_0790.html">Using HDFS</a></div>
|
|
</div>
|
|
</div>
|
|
|