<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:wfw="http://wellformedweb.org/CommentAPI/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	xmlns:slash="http://purl.org/rss/1.0/modules/slash/"
	>

<channel>
	<title>Everlinux's Blog &#187; Storage</title>
	<atom:link href="http://everlinux.com/blog/category/storage/feed" rel="self" type="application/rss+xml" />
	<link>http://everlinux.com/blog</link>
	<description>Sempre vivendo, aprendendo e blogando... :)</description>
	<lastBuildDate>Fri, 30 Jul 2010 18:13:16 +0000</lastBuildDate>
	<language>en</language>
	<sy:updatePeriod>hourly</sy:updatePeriod>
	<sy:updateFrequency>1</sy:updateFrequency>
	<generator>http://wordpress.org/?v=3.0</generator>
		<item>
		<title>Como aumentar seu cluster online (ESX+GFS)</title>
		<link>http://everlinux.com/blog/2009/01/22/como-aumentar-seu-cluster-online-esxgfs</link>
		<comments>http://everlinux.com/blog/2009/01/22/como-aumentar-seu-cluster-online-esxgfs#comments</comments>
		<pubDate>Thu, 22 Jan 2009 22:28:19 +0000</pubDate>
		<dc:creator>Tiago Cruz</dc:creator>
				<category><![CDATA[Cluster]]></category>
		<category><![CDATA[Dicas]]></category>
		<category><![CDATA[ESX]]></category>
		<category><![CDATA[High Availability]]></category>
		<category><![CDATA[Red Hat]]></category>
		<category><![CDATA[Storage]]></category>
		<category><![CDATA[VMware]]></category>
		<category><![CDATA[Virtualização]]></category>
		<category><![CDATA[GFS]]></category>
		<category><![CDATA[LVM]]></category>

		<guid isPermaLink="false">http://www.everlinux.com/blog/?p=219</guid>
		<description><![CDATA[Neste exemplo eu tenho um cluster em produção com a LUN compartilhada de 500 GB via GFS. Pretendo dobrar sua capacidade sem parada no sistema. 1-) Adicionar um &#8220;Mapped Raw LUN&#8221; no ESX, em Compatibility Mode &#8220;Physical&#8221;. A &#8220;SCSI Controller&#8221; também precisa ser &#8220;Physical&#8221;, e o disco RAW precisa estar nesta controller dedicada Ex: Hard [...]]]></description>
			<content:encoded><![CDATA[<p>Neste exemplo eu tenho um cluster em produção com a LUN compartilhada de 500 GB via GFS. Pretendo dobrar sua capacidade sem parada no sistema.</p>
<p><strong>1-)</strong> Adicionar um &#8220;Mapped Raw LUN&#8221; no ESX, em Compatibility Mode &#8220;Physical&#8221;.<br />
A &#8220;SCSI Controller&#8221; também precisa ser &#8220;Physical&#8221;, e o disco RAW precisa estar nesta controller dedicada</p>
<p>Ex:<br />
Hard Disk 1 = SCSI Controller 0 = SCSI (0:0)<br />
Hard Disk 2 = SCSI Controller 1 = SCSI (1:0)<br />
Hard Disk 2 = SCSI Controller 1 = SCSI (1:1)</p>
<p><strong>2-) </strong>Adicionar a LUN no Linux:</p>

<div class="wp_syntax"><div class="code"><pre class="bash" style="font-family:monospace;"><span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-02 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># cat /proc/scsi/scsi </span>
Attached devices:
Host: scsi0 Channel: 00 Id: 00 Lun: 00
  Vendor: VMware   Model: Virtual disk     Rev: <span style="color: #000000;">1.0</span> 
  Type:   Direct-Access                    ANSI SCSI revision: 02
Host: scsi1 Channel: 00 Id: 00 Lun: 00
  Vendor: IBM      Model: <span style="color: #000000;">1815</span>      FAStT  Rev: 0914
  Type:   Direct-Access                    ANSI SCSI revision: 05
&nbsp;
<span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-02 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># echo &quot;scsi add-single-device 1 0 1 0&quot; &gt; /proc/scsi/scsi</span>
&nbsp;
<span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-02 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># cat /proc/scsi/scsi </span>
Attached devices:
Host: scsi0 Channel: 00 Id: 00 Lun: 00
  Vendor: VMware   Model: Virtual disk     Rev: <span style="color: #000000;">1.0</span> 
  Type:   Direct-Access                    ANSI SCSI revision: 02
Host: scsi1 Channel: 00 Id: 00 Lun: 00
  Vendor: IBM      Model: <span style="color: #000000;">1815</span>      FAStT  Rev: 0914
  Type:   Direct-Access                    ANSI SCSI revision: 05
Host: scsi1 Channel: 00 Id: 01 Lun: 00
  Vendor: IBM      Model: <span style="color: #000000;">1815</span>      FAStT  Rev: 0914
  Type:   Direct-Access                    ANSI SCSI revision: 05</pre></div></div>

<p>Sendo: Host &#8211; Channel &#8211; ID &#8211; Lun<br />
(um reboot também funcionaria <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_smile.gif' alt=':-)' class='wp-smiley' />  )</p>
<p><strong>3-)</strong> Configurar o Cluster de LVM</p>
<p>3a-) Criando um Logical Volume</p>

<div class="wp_syntax"><div class="code"><pre class="bash" style="font-family:monospace;"><span style="color: #666666; font-style: italic;"># fdisk /dev/sdc (usar ID 8e = Linux LVM)</span>
<span style="color: #666666; font-style: italic;"># partprobe (em todas as maquinas do cluster)</span>
<span style="color: #666666; font-style: italic;"># lvmconf --enable-cluster</span>
<span style="color: #666666; font-style: italic;"># pvcreate /dev/sdb1</span>
<span style="color: #666666; font-style: italic;"># vgcreate -Ay -cy pre_homolog /dev/sdb1</span>
<span style="color: #666666; font-style: italic;"># lvcreate -L 499G -n hotsites pre_homolog</span></pre></div></div>

<p>ou</p>
<p>3b-) Aumentando um já existente</p>

<div class="wp_syntax"><div class="code"><pre class="bash" style="font-family:monospace;"><span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-01 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># pvs</span>
  PV         VG          Fmt  Attr PSize   PFree
  <span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>sdb1  pre_homolog lvm2 a-   499.99G 1016.00M
  <span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>sdc1              lvm2 <span style="color: #660033;">--</span>   499.99G  499.99G
&nbsp;
<span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-01 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># vgextend pre_homolog /dev/sdc1</span>
  Volume group <span style="color: #ff0000;">&quot;pre_homolog&quot;</span> successfully extended
&nbsp;
<span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-01 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># vgs</span>
  VG          <span style="color: #666666; font-style: italic;">#PV #LV #SN Attr   VSize   VFree  </span>
  pre_homolog   <span style="color: #000000;">2</span>   <span style="color: #000000;">1</span>   <span style="color: #000000;">0</span> wz--nc 999.98G 500.98G
&nbsp;
<span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-01 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># pvs</span>
  PV         VG          Fmt  Attr PSize   PFree
  <span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>sdb1  pre_homolog lvm2 a-   499.99G 1016.00M
  <span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>sdc1  pre_homolog lvm2 a-   499.99G  499.99G
&nbsp;
<span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-01 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># lvextend -Ay -l+128252 /dev/pre_homolog/hotsites</span>
  Extending logical volume hotsites to <span style="color: #000000;">999.98</span> GB
  Logical volume hotsites successfully resized
&nbsp;
<span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-01 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># lvs</span>
  LV       VG          Attr   LSize   Origin Snap<span style="color: #000000; font-weight: bold;">%</span>  Move Log Copy<span style="color: #000000; font-weight: bold;">%</span>  Convert
  hotsites pre_homolog <span style="color: #660033;">-wi-ao</span> 999.98G
&nbsp;
<span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-01 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># vgs</span>
  VG          <span style="color: #666666; font-style: italic;">#PV #LV #SN Attr   VSize   VFree</span>
  pre_homolog   <span style="color: #000000;">2</span>   <span style="color: #000000;">1</span>   <span style="color: #000000;">0</span> wz--nc 999.98G    <span style="color: #000000;">0</span>
&nbsp;
<span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-01 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># pvs</span>
  PV         VG          Fmt  Attr PSize   PFree
  <span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>sdb1  pre_homolog lvm2 a-   499.99G    <span style="color: #000000;">0</span>
  <span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>sdc1  pre_homolog lvm2 a-   499.99G    <span style="color: #000000;">0</span></pre></div></div>

<p><strong>4-)</strong> Aumentar o GFS</p>
<p>Com o /gfs montado, usando o gfs_grow</p>

<div class="wp_syntax"><div class="code"><pre class="bash" style="font-family:monospace;"><span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-01 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># gfs_grow -v /gfs</span>
FS: Mount Point: <span style="color: #000000; font-weight: bold;">/</span>gfs
FS: Device: <span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>mapper<span style="color: #000000; font-weight: bold;">/</span>pre_homolog-hotsites
FS: Options: rw,<span style="color: #007800;">hostdata</span>=<span style="color: #007800;">jid</span>=<span style="color: #000000;">0</span>:<span style="color: #007800;"><span style="color: #c20cb9; font-weight: bold;">id</span></span>=<span style="color: #000000;">131073</span>:<span style="color: #007800;">first</span>=<span style="color: #000000;">0</span>
FS: Size: <span style="color: #000000;">130809855</span>
RGRP: Current Resource Group List:
RI: Addr <span style="color: #000000;">130744354</span>, RgLen <span style="color: #000000;">5</span>, Start <span style="color: #000000;">130744359</span>, DataLen <span style="color: #000000;">65496</span>, BmapLen <span style="color: #000000;">16374</span>
RI: Addr <span style="color: #000000;">130678852</span>, RgLen <span style="color: #000000;">5</span>, Start <span style="color: #000000;">130678857</span>, DataLen <span style="color: #000000;">65496</span>, BmapLen <span style="color: #000000;">16374</span>
RI: Addr <span style="color: #000000;">130613350</span>, RgLen <span style="color: #000000;">5</span>, Start <span style="color: #000000;">130613355</span>, DataLen <span style="color: #000000;">65496</span>, BmapLen <span style="color: #000000;">16374</span>
...
...
RI: Addr <span style="color: #000000;">261895985</span>, RgLen <span style="color: #000000;">15</span>, Start <span style="color: #000000;">261896000</span>, DataLen <span style="color: #000000;">243904</span>, BmapLen <span style="color: #000000;">60976</span>
RGRP: <span style="color: #000000;">539</span> Resource <span style="color: #c20cb9; font-weight: bold;">groups</span> <span style="color: #000000; font-weight: bold;">in</span> total
Preparing to <span style="color: #c20cb9; font-weight: bold;">write</span> new FS information...
Done.</pre></div></div>

<p>Antes:</p>

<div class="wp_syntax"><div class="code"><pre class="bash" style="font-family:monospace;"><span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-02 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># df -h</span>
Filesystem            Size  Used Avail Use<span style="color: #000000; font-weight: bold;">%</span> Mounted on
<span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>sda1             6.9G  1.8G  4.7G  <span style="color: #000000;">28</span><span style="color: #000000; font-weight: bold;">%</span> <span style="color: #000000; font-weight: bold;">/</span>
tmpfs                 1.9G     <span style="color: #000000;">0</span>  1.9G   <span style="color: #000000;">0</span><span style="color: #000000; font-weight: bold;">%</span> <span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>shm
<span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>mapper<span style="color: #000000; font-weight: bold;">/</span>pre_homolog-hotsites
                      493G   12G  482G   <span style="color: #000000;">3</span><span style="color: #000000; font-weight: bold;">%</span> <span style="color: #000000; font-weight: bold;">/</span>gfs</pre></div></div>

<p>Depois:</p>

<div class="wp_syntax"><div class="code"><pre class="bash" style="font-family:monospace;"><span style="color: #7a0874; font-weight: bold;">&#91;</span>root<span style="color: #000000; font-weight: bold;">@</span>cluster-02 ~<span style="color: #7a0874; font-weight: bold;">&#93;</span><span style="color: #666666; font-style: italic;"># df -h</span>
Filesystem            Size  Used Avail Use<span style="color: #000000; font-weight: bold;">%</span> Mounted on
<span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>sda1             6.9G  1.8G  4.7G  <span style="color: #000000;">28</span><span style="color: #000000; font-weight: bold;">%</span> <span style="color: #000000; font-weight: bold;">/</span>
tmpfs                 1.9G     <span style="color: #000000;">0</span>  1.9G   <span style="color: #000000;">0</span><span style="color: #000000; font-weight: bold;">%</span> <span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>shm
<span style="color: #000000; font-weight: bold;">/</span>dev<span style="color: #000000; font-weight: bold;">/</span>mapper<span style="color: #000000; font-weight: bold;">/</span>pre_homolog-hotsites
                      994G   12G  983G   <span style="color: #000000;">2</span><span style="color: #000000; font-weight: bold;">%</span> <span style="color: #000000; font-weight: bold;">/</span>gfs</pre></div></div>

]]></content:encoded>
			<wfw:commentRss>http://everlinux.com/blog/2009/01/22/como-aumentar-seu-cluster-online-esxgfs/feed</wfw:commentRss>
		<slash:comments>2</slash:comments>
		</item>
		<item>
		<title>Cluster com VMware ESX + GFS</title>
		<link>http://everlinux.com/blog/2009/01/06/cluster-com-vmware-esx-gfs</link>
		<comments>http://everlinux.com/blog/2009/01/06/cluster-com-vmware-esx-gfs#comments</comments>
		<pubDate>Tue, 06 Jan 2009 23:26:47 +0000</pubDate>
		<dc:creator>Tiago Cruz</dc:creator>
				<category><![CDATA[Cluster]]></category>
		<category><![CDATA[Dicas]]></category>
		<category><![CDATA[ESX]]></category>
		<category><![CDATA[High Availability]]></category>
		<category><![CDATA[Linux]]></category>
		<category><![CDATA[Scripts]]></category>
		<category><![CDATA[Storage]]></category>
		<category><![CDATA[VMware]]></category>
		<category><![CDATA[Virtualização]]></category>
		<category><![CDATA[GFS]]></category>

		<guid isPermaLink="false">http://www.everlinux.com/blog/2009/01/06/cluster-com-vmware-esx-gfs/</guid>
		<description><![CDATA[Já escrevemos bastante sobre Virtualização usando Xen e Cluster aqui no Blog. Desta vez, para começar bem 2009 eu irei apenas citar alguns testes que fiz usando VMware ESX com GFS, em um Storage compartilhado, usando tecnologia Red Hat 5.2 x86_64 com um Fence_VMWare Tech Preview do RHEL 5.3 beta. O Problema Eu já o [...]]]></description>
			<content:encoded><![CDATA[<p><a href='http://www.everlinux.com/blog/wp-content/uploads/2009/01/teste_esx_gfs_ever.png' title='Teste_ESX_GFS_Ever'><img src='http://www.everlinux.com/blog/wp-content/uploads/2009/01/teste_esx_gfs_ever.thumbnail.png' alt='Teste_ESX_GFS_Ever' /></a></p>
<p>Já escrevemos bastante sobre <a href="http://www.everlinux.com/blog/category/virtualizacao/">Virtualização </a>usando <a href="http://www.everlinux.com/blog/category/xen/">Xen</a> e <a href="http://www.everlinux.com/blog/category/cluster/">Cluster</a> aqui no Blog. </p>
<p>Desta vez, para começar bem 2009 eu irei apenas citar alguns testes que fiz usando VMware ESX com GFS, em um <a href="http://www.everlinux.com/blog/category/storage/">Storage</a> compartilhado, usando tecnologia Red Hat 5.2 x86_64 com um Fence_VMWare Tech Preview do RHEL 5.3 beta.</p>
<p><strong>O Problema</strong></p>
<p>Eu já o utilizava o GFS com Xen ou até mesmo com <a href="http://www.everlinux.com/blog/2008/06/16/usando-drbd-com-gfs-e-fencing-manual/">DRBD em Muti-Master</a>, porém o grande problema é que não existia um fence para vmware que funcionasse a contento.</p>
<p>O <a href="http://sources.redhat.com/cgi-bin/cvsweb.cgi/cluster/fence/agents/vmware/fence_vmware.pl?cvsroot=cluster">fence_vmware</a>, escrito em perl parou no tempo em Agosto de 2007 e fora que a <a href="http://www.vmware.com/download/sdk/api.html">VMware Perl API</a> necessária não compila em 64 bits.</p>
<p>A Red Hat pretende lançar agora no RHEL 5.3 um <a href="http://sources.redhat.com/cluster/wiki/VMware_FencingConfig">fence_vmware re-escrito em Python</a>, porém ele ainda não trabalha com Cluster, pois conecta-se direto na máquina física, e não no VMCenter.</p>
<p><strong>A Solução</strong></p>
<p>A solução é utilizar o vmware_fence_vi, também da Red Hat que está em fase de desenvolvimento atualmente, que conecta no Virtual Center e não na máquina física que pode estar morta <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_smile.gif' alt=':)' class='wp-smiley' /> </p>
<p>Segue o /etc/cluster/cluster.conf utilizado nos testes:</p>
<p><strong>Configuração do Cluster</strong></p>

<div class="wp_syntax"><div class="code"><pre class="bash" style="font-family:monospace;"><span style="color: #000000; font-weight: bold;">&lt;</span>cluster <span style="color: #007800;"><span style="color: #7a0874; font-weight: bold;">alias</span></span>=<span style="color: #ff0000;">&quot;Cluster_GFS&quot;</span> <span style="color: #007800;">config_version</span>=<span style="color: #ff0000;">&quot;10&quot;</span> <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;Cluster_GFS&quot;</span><span style="color: #000000; font-weight: bold;">&gt;</span>
	<span style="color: #000000; font-weight: bold;">&lt;</span>fence_daemon <span style="color: #007800;">clean_start</span>=<span style="color: #ff0000;">&quot;0&quot;</span> <span style="color: #007800;">post_fail_delay</span>=<span style="color: #ff0000;">&quot;0&quot;</span> <span style="color: #007800;">post_join_delay</span>=<span style="color: #ff0000;">&quot;3&quot;</span><span style="color: #000000; font-weight: bold;">/&gt;</span>
<span style="color: #000000; font-weight: bold;">&lt;</span>cman<span style="color: #000000; font-weight: bold;">&gt;</span>
    <span style="color: #000000; font-weight: bold;">&lt;</span>multicast <span style="color: #007800;">addr</span>=<span style="color: #ff0000;">&quot;224.0.0.1&quot;</span> <span style="color: #007800;">interface</span>=<span style="color: #ff0000;">&quot;eth1&quot;</span><span style="color: #000000; font-weight: bold;">/&gt;</span>
<span style="color: #000000; font-weight: bold;">&lt;/</span>cman<span style="color: #000000; font-weight: bold;">&gt;</span>
	<span style="color: #000000; font-weight: bold;">&lt;</span>clusternodes<span style="color: #000000; font-weight: bold;">&gt;</span>
		<span style="color: #000000; font-weight: bold;">&lt;</span>clusternode <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;cluster-01-node-01&quot;</span> <span style="color: #007800;">nodeid</span>=<span style="color: #ff0000;">&quot;1&quot;</span> <span style="color: #007800;">votes</span>=<span style="color: #ff0000;">&quot;1&quot;</span><span style="color: #000000; font-weight: bold;">&gt;</span>
			<span style="color: #000000; font-weight: bold;">&lt;</span>multicast <span style="color: #007800;">addr</span>=<span style="color: #ff0000;">&quot;224.0.0.1&quot;</span> <span style="color: #007800;">interface</span>=<span style="color: #ff0000;">&quot;eth1&quot;</span><span style="color: #000000; font-weight: bold;">/&gt;</span>
	           <span style="color: #000000; font-weight: bold;">&lt;</span>fence<span style="color: #000000; font-weight: bold;">&gt;</span>
        	        <span style="color: #000000; font-weight: bold;">&lt;</span>method <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;single&quot;</span><span style="color: #000000; font-weight: bold;">&gt;</span>
                	        <span style="color: #000000; font-weight: bold;">&lt;</span>device <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;node-01&quot;</span><span style="color: #000000; font-weight: bold;">/&gt;</span>
	                <span style="color: #000000; font-weight: bold;">&lt;/</span>method<span style="color: #000000; font-weight: bold;">&gt;</span>
	           <span style="color: #000000; font-weight: bold;">&lt;/</span>fence<span style="color: #000000; font-weight: bold;">&gt;</span>
                <span style="color: #000000; font-weight: bold;">&lt;/</span>clusternode<span style="color: #000000; font-weight: bold;">&gt;</span>
                <span style="color: #000000; font-weight: bold;">&lt;</span>clusternode <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;cluster-01-node-02&quot;</span> <span style="color: #007800;">nodeid</span>=<span style="color: #ff0000;">&quot;2&quot;</span> <span style="color: #007800;">votes</span>=<span style="color: #ff0000;">&quot;1&quot;</span><span style="color: #000000; font-weight: bold;">&gt;</span>
			<span style="color: #000000; font-weight: bold;">&lt;</span>multicast <span style="color: #007800;">addr</span>=<span style="color: #ff0000;">&quot;224.0.0.1&quot;</span> <span style="color: #007800;">interface</span>=<span style="color: #ff0000;">&quot;eth1&quot;</span><span style="color: #000000; font-weight: bold;">/&gt;</span>
                   <span style="color: #000000; font-weight: bold;">&lt;</span>fence<span style="color: #000000; font-weight: bold;">&gt;</span>
                        <span style="color: #000000; font-weight: bold;">&lt;</span>method <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;single&quot;</span><span style="color: #000000; font-weight: bold;">&gt;</span>
                                <span style="color: #000000; font-weight: bold;">&lt;</span>device <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;node-02&quot;</span><span style="color: #000000; font-weight: bold;">/&gt;</span>
                        <span style="color: #000000; font-weight: bold;">&lt;/</span>method<span style="color: #000000; font-weight: bold;">&gt;</span>
                   <span style="color: #000000; font-weight: bold;">&lt;/</span>fence<span style="color: #000000; font-weight: bold;">&gt;</span>
                <span style="color: #000000; font-weight: bold;">&lt;/</span>clusternode<span style="color: #000000; font-weight: bold;">&gt;</span>
                <span style="color: #000000; font-weight: bold;">&lt;</span>clusternode <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;cluster-01-node-03&quot;</span> <span style="color: #007800;">nodeid</span>=<span style="color: #ff0000;">&quot;3&quot;</span> <span style="color: #007800;">votes</span>=<span style="color: #ff0000;">&quot;1&quot;</span><span style="color: #000000; font-weight: bold;">&gt;</span>
			<span style="color: #000000; font-weight: bold;">&lt;</span>multicast <span style="color: #007800;">addr</span>=<span style="color: #ff0000;">&quot;224.0.0.1&quot;</span> <span style="color: #007800;">interface</span>=<span style="color: #ff0000;">&quot;eth1&quot;</span><span style="color: #000000; font-weight: bold;">/&gt;</span>
                   <span style="color: #000000; font-weight: bold;">&lt;</span>fence<span style="color: #000000; font-weight: bold;">&gt;</span>
                        <span style="color: #000000; font-weight: bold;">&lt;</span>method <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;single&quot;</span><span style="color: #000000; font-weight: bold;">&gt;</span>
                                <span style="color: #000000; font-weight: bold;">&lt;</span>device <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;node-03&quot;</span><span style="color: #000000; font-weight: bold;">/&gt;</span>
                        <span style="color: #000000; font-weight: bold;">&lt;/</span>method<span style="color: #000000; font-weight: bold;">&gt;</span>
                   <span style="color: #000000; font-weight: bold;">&lt;/</span>fence<span style="color: #000000; font-weight: bold;">&gt;</span>
                <span style="color: #000000; font-weight: bold;">&lt;/</span>clusternode<span style="color: #000000; font-weight: bold;">&gt;</span>
	<span style="color: #000000; font-weight: bold;">&lt;/</span>clusternodes<span style="color: #000000; font-weight: bold;">&gt;</span>
	<span style="color: #000000; font-weight: bold;">&lt;</span>fencedevices<span style="color: #000000; font-weight: bold;">&gt;</span>
		<span style="color: #000000; font-weight: bold;">&lt;</span>fencedevice <span style="color: #007800;">agent</span>=<span style="color: #ff0000;">&quot;fence_vmware_vi&quot;</span> <span style="color: #007800;">ipaddr</span>=<span style="color: #ff0000;">&quot;10.107.113.2&quot;</span> <span style="color: #007800;"><span style="color: #c20cb9; font-weight: bold;">login</span></span>=<span style="color: #ff0000;">&quot;user&quot;</span> <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;node-01&quot;</span> <span style="color: #007800;"><span style="color: #c20cb9; font-weight: bold;">passwd</span></span>=<span style="color: #ff0000;">&quot;xxxxx&quot;</span> <span style="color: #007800;">port</span>=<span style="color: #ff0000;">&quot;Cluster_01_Node_01&quot;</span><span style="color: #000000; font-weight: bold;">/&gt;</span>
		<span style="color: #000000; font-weight: bold;">&lt;</span>fencedevice <span style="color: #007800;">agent</span>=<span style="color: #ff0000;">&quot;fence_vmware_vi&quot;</span> <span style="color: #007800;">ipaddr</span>=<span style="color: #ff0000;">&quot;10.107.113.2&quot;</span> <span style="color: #007800;"><span style="color: #c20cb9; font-weight: bold;">login</span></span>=<span style="color: #ff0000;">&quot;user&quot;</span> <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;node-02&quot;</span> <span style="color: #007800;"><span style="color: #c20cb9; font-weight: bold;">passwd</span></span>=<span style="color: #ff0000;">&quot;xxxxx&quot;</span> <span style="color: #007800;">port</span>=<span style="color: #ff0000;">&quot;Cluster_01_Node_02&quot;</span><span style="color: #000000; font-weight: bold;">/&gt;</span>
		<span style="color: #000000; font-weight: bold;">&lt;</span>fencedevice <span style="color: #007800;">agent</span>=<span style="color: #ff0000;">&quot;fence_vmware_vi&quot;</span> <span style="color: #007800;">ipaddr</span>=<span style="color: #ff0000;">&quot;10.107.113.2&quot;</span> <span style="color: #007800;"><span style="color: #c20cb9; font-weight: bold;">login</span></span>=<span style="color: #ff0000;">&quot;user&quot;</span> <span style="color: #007800;">name</span>=<span style="color: #ff0000;">&quot;node-03&quot;</span> <span style="color: #007800;"><span style="color: #c20cb9; font-weight: bold;">passwd</span></span>=<span style="color: #ff0000;">&quot;xxxxx&quot;</span> <span style="color: #007800;">port</span>=<span style="color: #ff0000;">&quot;Cluster_01_Node_03&quot;</span><span style="color: #000000; font-weight: bold;">/&gt;</span>
	<span style="color: #000000; font-weight: bold;">&lt;/</span>fencedevices<span style="color: #000000; font-weight: bold;">&gt;</span>
<span style="color: #000000; font-weight: bold;">&lt;/</span>cluster<span style="color: #000000; font-weight: bold;">&gt;</span></pre></div></div>

<p>Notas: </p>
<p>1-) Troque parênteses por sinal de maior/ menor (maldito wordpress!) <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_sad.gif' alt=':(' class='wp-smiley' /> </p>
<p>2-) O IP 10.107.113.2 é o IP do Virtual Center. Ele sempre sabe onde a máquina virtual se encontra, mesmo que ela for migrada para outro host físico.</p>
<p>3-) Forçamos o MultiCast pela eth1, que é uma interface dedicada para o heartbeat do Cluster. Algumas vezes o multicast tentava sair pela interface de loopback e cada máquina achava que estava sozinha no cluster <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_sad.gif' alt=':(' class='wp-smiley' /> </p>
<p>/etc/hosts dos servidores:</p>
<p><code>10.0.0.1	cluster-01-node-01<br />
10.0.0.2	cluster-01-node-02<br />
10.0.0.3	cluster-01-node-03<br />
10.0.1.1	node-01<br />
10.0.1.2	node-02<br />
10.0.1.3	node-03</code></p>
<p><strong>Script de escrita</strong></p>
<p>Foi utilizado um script para escrever no storage compartilhado a cada segundo com o nome da máquina e a hora exata (minuto, segundo) da escrita.</p>
<p><code># for i in `seq 1 1000`; do a=`echo $RANDOM | cut -c1-2`; dd if=/dev/zero of=/teste/`hostname -s`-`date +%H_%M_%S` bs=512k count=$a; sleep 1; done</code></p>
<p><strong>Nós do Cluster</strong></p>
<p><code>[root@cluster-01-node-02 ~]# cman_tool nodes<br />
Node  Sts   Inc   Joined               Name<br />
   1   M     60   2009-01-06 19:46:11  cluster-01-node-01<br />
   2   M     56   2009-01-06 19:46:11  cluster-01-node-02<br />
   3   M     68   2009-01-06 19:56:35  cluster-01-node-03</code></p>
<p><strong>Fence Device</strong></p>

<div class="wp_syntax"><div class="code"><pre class="bash" style="font-family:monospace;"><span style="color: #000000; font-weight: bold;">/</span>sbin<span style="color: #000000; font-weight: bold;">/</span>fence_vmware_vi - DEVEL.1229423489
&nbsp;
<span style="color: #666666; font-style: italic;"># The Following agent has been tested on:</span>
<span style="color: #666666; font-style: italic;"># VI Perl API 1.6 against:</span>
<span style="color: #666666; font-style: italic;">#       VMware ESX 3.5</span>
<span style="color: #666666; font-style: italic;">#       VMware ESXi 3.5 update 2</span>
<span style="color: #666666; font-style: italic;">#       VMware Virtual Center 2.5</span>
<span style="color: #666666; font-style: italic;">#&lt;/code&gt;</span></pre></div></div>

<p>Com a seguinte modificação na linha 13:</p>

<div class="wp_syntax"><div class="code"><pre class="bash" style="font-family:monospace;"><span style="color: #666666; font-style: italic;">#sys.path.append(&quot;@FENCEAGENTSLIBDIR@&quot;)</span>
sys.path.append<span style="color: #7a0874; font-weight: bold;">&#40;</span><span style="color: #ff0000;">&quot;/usr/lib/fence&quot;</span><span style="color: #7a0874; font-weight: bold;">&#41;</span></pre></div></div>

<p><strong>Desligar máquina VIRTUAL</strong></p>
<p>Teste: Desligar máquina VIRTUAL cluster-01-node-03</p>
<p>Resultado: GFS congelado por ~1 minuto</p>
<p>O que ocorreu: O fence_vmware_vi conectou no vmcenter e tirou a máquina do Cluster, liberando assim o I/O.</p>
<p><strong>Desligar máquina FISICA normalmente</strong></p>
<p>Teste: Desligar com reboot máquina FISICA 01 (que tinha a cluster-01-node-03)</p>
<p>Resultado: GFS congelado por ~2 minutos</p>
<p>O que ocorreu: Máquinas física iniciou o processo de reboot, a máquina virtual cluster-01-node-03 morreu e foi migrada e reiniciada em outro host físico.</p>
<p>PS: A máquina virtual não retornou ao seu host de origem quando a máquina física voltou</p>
<p><code>-rw-r--r-- 1 root root 44040192 Jan  6  2009 cluster-01-node-01-19_50_52<br />
-rw-r--r-- 1 root root 15204352 Jan  6  2009 cluster-01-node-01-19_50_53<br />
-rw-r--r-- 1 root root 10485760 Jan  6  2009 cluster-01-node-01-19_50_54<br />
-rw-r--r-- 1 root root 41418752 Jan  6  2009 cluster-01-node-01-19_52_57<br />
-rw-r--r-- 1 root root 17825792 Jan  6  2009 cluster-01-node-01-19_52_58<br />
-rw-r--r-- 1 root root 23068672 Jan  6  2009 cluster-01-node-01-19_52_59<br />
-rw-r--r-- 1 root root  6815744 Jan  6  2009 cluster-01-node-01-19_53_00<br />
-rw-r--r-- 1 root root 49807360 Jan  6  2009 cluster-01-node-01-19_53_01<br />
-rw-r--r-- 1 root root 10485760 Jan  6  2009 cluster-01-node-01-19_53_03<br />
</code></p>
<p><strong>Desligar máquina FISICA brutalmente</strong></p>
<p>Desligar com RESET (via RSA) máquina FISICA 02 (que tinha a cluster-01-node-03)</p>
<p>Resultado: GFS congelado por ~3 minutos</p>
<p>O que ocorreu: Máquinas física foi brutalmente desligada, a máquina virtual cluster-01-node-03 morreu e foi migrada e reiniciada em outro host físico.</p>
<p>O fence_vmware_vi tentou várias vezes tirar a máquina do cluster com falhas consecutivas pois a máquina não mais existia e o ESX demorou um pouco para entender o que estava acontecendo.</p>
<p>PS: A máquina virtual não retornou ao seu host de origem quando a máquina física voltou</p>
<p><code>-rw-r--r-- 1 root root 12058624 Jan  6 20:17 cluster-01-node-01-20_17_57<br />
-rw-r--r-- 1 root root 11534336 Jan  6 20:17 cluster-01-node-01-20_17_58<br />
-rw-r--r-- 1 root root 10485760 Jan  6 20:17 cluster-01-node-01-20_17_59<br />
-rw-r--r-- 1 root root  5242880 Jan  6 20:18 cluster-01-node-01-20_18_00<br />
-rw-r--r-- 1 root root  5767168 Jan  6 20:20 cluster-01-node-01-20_18_01<br />
-rw-r--r-- 1 root root  6291456 Jan  6 20:21 cluster-01-node-01-20_21_00<br />
-rw-r--r-- 1 root root 37748736 Jan  6 20:21 cluster-01-node-01-20_21_02<br />
-rw-r--r-- 1 root root 17825792 Jan  6 20:21 cluster-01-node-01-20_21_03<br />
-rw-r--r-- 1 root root 12582912 Jan  6 20:21 cluster-01-node-01-20_21_04<br />
-rw-r--r-- 1 root root 34603008 Jan  6 20:21 cluster-01-node-01-20_21_05<br />
-rw-r--r-- 1 root root  5242880 Jan  6 20:21 cluster-01-node-01-20_21_06<br />
-rw-r--r-- 1 root root 12582912 Jan  6 20:21 cluster-01-node-01-20_21_07<br />
-rw-r--r-- 1 root root 12582912 Jan  6 20:21 cluster-01-node-01-20_21_08</code></p>
<p><strong>Trechos de logs</strong></p>

<div class="wp_syntax"><div class="code"><pre class="bash" style="font-family:monospace;">Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">10</span> cluster-01-node-01 openais<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2646</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: <span style="color: #7a0874; font-weight: bold;">&#91;</span>TOTEM<span style="color: #7a0874; font-weight: bold;">&#93;</span> The token was lost <span style="color: #000000; font-weight: bold;">in</span> the OPERATIONAL state. 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">10</span> cluster-01-node-01 openais<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2646</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: <span style="color: #7a0874; font-weight: bold;">&#91;</span>TOTEM<span style="color: #7a0874; font-weight: bold;">&#93;</span> Receive multicast socket recv buffer <span style="color: #c20cb9; font-weight: bold;">size</span> <span style="color: #7a0874; font-weight: bold;">&#40;</span><span style="color: #000000;">288000</span> bytes<span style="color: #7a0874; font-weight: bold;">&#41;</span>. 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">10</span> cluster-01-node-01 openais<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2646</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: <span style="color: #7a0874; font-weight: bold;">&#91;</span>TOTEM<span style="color: #7a0874; font-weight: bold;">&#93;</span> Transmit multicast socket send buffer <span style="color: #c20cb9; font-weight: bold;">size</span> <span style="color: #7a0874; font-weight: bold;">&#40;</span><span style="color: #000000;">262142</span> bytes<span style="color: #7a0874; font-weight: bold;">&#41;</span>. 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">10</span> cluster-01-node-01 openais<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2646</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: <span style="color: #7a0874; font-weight: bold;">&#91;</span>TOTEM<span style="color: #7a0874; font-weight: bold;">&#93;</span> entering GATHER state from 2. 
&nbsp;
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">15</span> cluster-01-node-01 openais<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2646</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: <span style="color: #7a0874; font-weight: bold;">&#91;</span>CLM  <span style="color: #7a0874; font-weight: bold;">&#93;</span> CLM CONFIGURATION CHANGE 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">15</span> cluster-01-node-01 openais<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2646</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: <span style="color: #7a0874; font-weight: bold;">&#91;</span>CLM  <span style="color: #7a0874; font-weight: bold;">&#93;</span> New Configuration: 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">15</span> cluster-01-node-01 openais<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2646</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: <span style="color: #7a0874; font-weight: bold;">&#91;</span>CLM  <span style="color: #7a0874; font-weight: bold;">&#93;</span> 	r<span style="color: #7a0874; font-weight: bold;">&#40;</span><span style="color: #000000;">0</span><span style="color: #7a0874; font-weight: bold;">&#41;</span> ip<span style="color: #7a0874; font-weight: bold;">&#40;</span>10.0.0.1<span style="color: #7a0874; font-weight: bold;">&#41;</span>  
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">15</span> cluster-01-node-01 openais<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2646</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: <span style="color: #7a0874; font-weight: bold;">&#91;</span>CLM  <span style="color: #7a0874; font-weight: bold;">&#93;</span> 	r<span style="color: #7a0874; font-weight: bold;">&#40;</span><span style="color: #000000;">0</span><span style="color: #7a0874; font-weight: bold;">&#41;</span> ip<span style="color: #7a0874; font-weight: bold;">&#40;</span>10.0.0.2<span style="color: #7a0874; font-weight: bold;">&#41;</span>  
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">15</span> cluster-01-node-01 openais<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2646</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: <span style="color: #7a0874; font-weight: bold;">&#91;</span>CLM  <span style="color: #7a0874; font-weight: bold;">&#93;</span> Members Left: 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">15</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: cluster-01-node-03 not a cluster member after <span style="color: #000000;">0</span> sec post_fail_delay
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">15</span> cluster-01-node-01 openais<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2646</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: <span style="color: #7a0874; font-weight: bold;">&#91;</span>CLM  <span style="color: #7a0874; font-weight: bold;">&#93;</span> 	r<span style="color: #7a0874; font-weight: bold;">&#40;</span><span style="color: #000000;">0</span><span style="color: #7a0874; font-weight: bold;">&#41;</span> ip<span style="color: #7a0874; font-weight: bold;">&#40;</span>10.0.0.3<span style="color: #7a0874; font-weight: bold;">&#41;</span>  
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">15</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: fencing node <span style="color: #ff0000;">&quot;cluster-01-node-03&quot;</span>
&nbsp;
  Please use <span style="color: #ff0000;">'-h'</span> <span style="color: #000000; font-weight: bold;">for</span> <span style="color: #007800;">usagcation</span>=HASH<span style="color: #7a0874; font-weight: bold;">&#40;</span>0x1df33bd0<span style="color: #7a0874; font-weight: bold;">&#41;</span>ing with the remote host.eports: vmware_helper returned Cannot power off vm Cluster_01_Node_03<span style="color: #000000; font-weight: bold;">!</span>
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">47</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: agent <span style="color: #ff0000;">&quot;fence_vmware_vi&quot;</span> reports: e 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">47</span> cluster-01-node-01 ccsd<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2636</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: Attempt to close an unopened CCS descriptor <span style="color: #7a0874; font-weight: bold;">&#40;</span><span style="color: #000000;">44880</span><span style="color: #7a0874; font-weight: bold;">&#41;</span>. 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">47</span> cluster-01-node-01 ccsd<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2636</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: Error <span style="color: #000000; font-weight: bold;">while</span> processing disconnect: Invalid request descriptor 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">47</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: fence <span style="color: #ff0000;">&quot;cluster-01-node-03&quot;</span> failed
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">18</span>:<span style="color: #000000;">52</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: fencing node <span style="color: #ff0000;">&quot;cluster-01-node&quot;</span>
&nbsp;
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">19</span>:<span style="color: #000000;">28</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: agent <span style="color: #ff0000;">&quot;fence_vmware_vi&quot;</span> reports: Connection timed out 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">19</span>:<span style="color: #000000;">28</span> cluster-01-node-01 ccsd<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2636</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: Attempt to close an unopened CCS descriptor <span style="color: #7a0874; font-weight: bold;">&#40;</span><span style="color: #000000;">45600</span><span style="color: #7a0874; font-weight: bold;">&#41;</span>. 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">19</span>:<span style="color: #000000;">28</span> cluster-01-node-01 ccsd<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2636</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: Error <span style="color: #000000; font-weight: bold;">while</span> processing disconnect: Invalid request descriptor 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">19</span>:<span style="color: #000000;">28</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: fence <span style="color: #ff0000;">&quot;cluster-01-node-03&quot;</span> failed
&nbsp;
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">19</span>:<span style="color: #000000;">33</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: fencing node <span style="color: #ff0000;">&quot;cluster-01-node-03&quot;</span>
  Please use <span style="color: #ff0000;">'-h'</span> <span style="color: #007800;">tNotConnected</span>=HASH<span style="color: #7a0874; font-weight: bold;">&#40;</span>0x19bb0090<span style="color: #7a0874; font-weight: bold;">&#41;</span>mote host, since it is disconnected.ware_helper returned Cannot power on vm Cluster_01_Node_03<span style="color: #000000; font-weight: bold;">!</span>
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">19</span>:<span style="color: #000000;">46</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: agent <span style="color: #ff0000;">&quot;fence_vmware_vi&quot;</span> reports: <span style="color: #000000; font-weight: bold;">for</span> usage 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">19</span>:<span style="color: #000000;">46</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: fence <span style="color: #ff0000;">&quot;cluster-01-node-03&quot;</span> failed
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">19</span>:<span style="color: #000000;">51</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: fencing node <span style="color: #ff0000;">&quot;cluster-01-node-03&quot;</span>
&nbsp;
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">20</span>:<span style="color: #000000;">59</span> cluster-01-node-01 ccsd<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2636</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: Attempt to close an unopened CCS descriptor <span style="color: #7a0874; font-weight: bold;">&#40;</span><span style="color: #000000;">46710</span><span style="color: #7a0874; font-weight: bold;">&#41;</span>. 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">20</span>:<span style="color: #000000;">59</span> cluster-01-node-01 ccsd<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2636</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: Error <span style="color: #000000; font-weight: bold;">while</span> processing disconnect: Invalid request descriptor 
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">20</span>:<span style="color: #000000;">59</span> cluster-01-node-01 fenced<span style="color: #7a0874; font-weight: bold;">&#91;</span><span style="color: #000000;">2666</span><span style="color: #7a0874; font-weight: bold;">&#93;</span>: fence <span style="color: #ff0000;">&quot;cluster-01-node-03.ig.com.br&quot;</span> success
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">20</span>:<span style="color: #000000;">59</span> cluster-01-node-01 kernel: GFS: <span style="color: #007800;">fsid</span>=Cluster_GFS:gfs.0: <span style="color: #007800;">jid</span>=<span style="color: #000000;">2</span>: Trying to acquire journal lock...
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">20</span>:<span style="color: #000000;">59</span> cluster-01-node-01 kernel: GFS: <span style="color: #007800;">fsid</span>=Cluster_GFS:gfs.0: <span style="color: #007800;">jid</span>=<span style="color: #000000;">2</span>: Looking at journal...
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">20</span>:<span style="color: #000000;">59</span> cluster-01-node-01 kernel: GFS: <span style="color: #007800;">fsid</span>=Cluster_GFS:gfs.0: <span style="color: #007800;">jid</span>=<span style="color: #000000;">2</span>: Acquiring the transaction lock...
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">20</span>:<span style="color: #000000;">59</span> cluster-01-node-01 kernel: GFS: <span style="color: #007800;">fsid</span>=Cluster_GFS:gfs.0: <span style="color: #007800;">jid</span>=<span style="color: #000000;">2</span>: Replaying journal...
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">20</span>:<span style="color: #000000;">59</span> cluster-01-node-01 kernel: GFS: <span style="color: #007800;">fsid</span>=Cluster_GFS:gfs.0: <span style="color: #007800;">jid</span>=<span style="color: #000000;">2</span>: Replayed <span style="color: #000000;">79</span> of <span style="color: #000000;">93</span> blocks
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">20</span>:<span style="color: #000000;">59</span> cluster-01-node-01 kernel: GFS: <span style="color: #007800;">fsid</span>=Cluster_GFS:gfs.0: <span style="color: #007800;">jid</span>=<span style="color: #000000;">2</span>: replays = <span style="color: #000000;">79</span>, skips = <span style="color: #000000;">13</span>, sames = <span style="color: #000000;">1</span>
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">20</span>:<span style="color: #000000;">59</span> cluster-01-node-01 kernel: GFS: <span style="color: #007800;">fsid</span>=Cluster_GFS:gfs.0: <span style="color: #007800;">jid</span>=<span style="color: #000000;">2</span>: Journal replayed <span style="color: #000000; font-weight: bold;">in</span> 1s
Jan  <span style="color: #000000;">6</span> <span style="color: #000000;">20</span>:<span style="color: #000000;">20</span>:<span style="color: #000000;">59</span> cluster-01-node-01 kernel: GFS: <span style="color: #007800;">fsid</span>=Cluster_GFS:gfs.0: <span style="color: #007800;">jid</span>=<span style="color: #000000;">2</span>: Done</pre></div></div>

<p><strong>Conclusão</strong></p>
<p>Estes testes foram realizados buscando alternativas que reduzam o downtime dos serviços e/ou aplicações contidos em um servidor, chegando a manter sua disponibilidade próxima dos 99,999%.</p>
<p>Como você pôde perceber, esta estrutura permite chegarmos a este número, mas até o momento em que está sendo escrito este post, a versão usada ainda não está homologada, usando pacotes ainda em desenvolvimento (cman+fence_vmware). </p>
<p>Num futuro próximo, já devemos ter algum retorno das empresas desenvolvedoras, pois focam em uma fatia do merdado muito forte e competitiva ($$$).</p>
]]></content:encoded>
			<wfw:commentRss>http://everlinux.com/blog/2009/01/06/cluster-com-vmware-esx-gfs/feed</wfw:commentRss>
		<slash:comments>3</slash:comments>
		</item>
		<item>
		<title>VMware Virtualization Forum 2008 &#8211; Eu fui!</title>
		<link>http://everlinux.com/blog/2008/06/19/vmware-virtualization-forum-2008-eu-fui</link>
		<comments>http://everlinux.com/blog/2008/06/19/vmware-virtualization-forum-2008-eu-fui#comments</comments>
		<pubDate>Fri, 20 Jun 2008 02:39:39 +0000</pubDate>
		<dc:creator>Tiago Cruz</dc:creator>
				<category><![CDATA[Eventos]]></category>
		<category><![CDATA[Linux]]></category>
		<category><![CDATA[Storage]]></category>
		<category><![CDATA[Virtualização]]></category>
		<category><![CDATA[Wii]]></category>

		<guid isPermaLink="false">http://www.everlinux.com/blog/2008/06/19/vmware-virtualization-forum-2008-eu-fui/</guid>
		<description><![CDATA[Participei hoje do evento &#8220;VMware Virtualization Forum 2008&#8220;, patrocinado pela VmWare e suas patrocinadoras como a Intel, HP, CA, EMC&#8230; Embora o evento maior da VmWare se passe &#8220;lá fora&#8221;, o evento aqui no Brasil foi bem interessante e tiveram suas inscrições esgotadas. O evento foi gratuíto, contando com café da manhã, Coffee Break, Almoço [...]]]></description>
			<content:encoded><![CDATA[<p>Participei hoje do evento &#8220;<a href="http://info.vmware.com/content/VirtualizationForum_Home">VMware Virtualization Forum 2008</a>&#8220;, patrocinado pela VmWare e suas patrocinadoras como a Intel, HP, CA, EMC&#8230; Embora o evento maior da VmWare se passe &#8220;lá fora&#8221;,  o evento aqui no Brasil foi bem interessante e tiveram <a href="http://info.vmware.com/content/PortgForum_Register?src=undefined&#038;ossrc=undefined">suas inscrições esgotadas</a>.</p>
<p><a href='http://www.everlinux.com/blog/wp-content/uploads/2008/06/nintendo_wii.jpg' title='Nintendo Wii'><img src='http://www.everlinux.com/blog/wp-content/uploads/2008/06/nintendo_wii.thumbnail.jpg' alt='Nintendo Wii' /></a></p>
<p>O evento foi gratuíto, contando com café da manhã, Coffee Break, Almoço e Coquetel de encerramento com o sorteio de alguns prêmios, como dois Ipod Touch e dois <a href="http://www.magazineluiza.com.br/especiais/nitendo/index.asp">Nintendos WII</a>, no qual eu tive a puta sorte de ganhar!!! <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_biggrin.gif' alt=':-D' class='wp-smiley' />  Nunca ganha nada em sorteio algum, porém dessa vez eu realmente preciso agradecer a Deus pelo presentão de aniversário que eu não teria condições de comprar!!! <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_biggrin.gif' alt=':-D' class='wp-smiley' />  Alias, agradeço de algum usuário desta maravilhosa tecnologia pudesse deixar alguns dicas para mim a respeito do console <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_wink.gif' alt=';)' class='wp-smiley' /> </p>
<p>Bom, voltando ao evento em si, com tanta coisa &#8220;de graça&#8221; bancada pelos patrocinadores, era de se esperar que os mesmos tentassem vender seus produtos durante as palestras. Não gostei de ver apresentações com slides &#8220;shareados&#8221; dando aquela sensação estranha Deja&#8217;vu&#8230;</p>
<p>Portando, com as palestras mais voltadas ao publico gerencial e não técnico, o evento deve ter agradado as pessoas que não precisam saber como o motor funciona, querem apenas ligar a chave e esperam que o carro saiam andando &#8212; e seus técnicos que se virem para aprender como a coisa funciona <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_smile.gif' alt=':)' class='wp-smiley' /> </p>
<p>Na parte da manhã tivemos palestras muito interessantes como:</p>
<p>-  a Polícia Federal, mostrando sua solução de Virtualização usando Blades (edizendo que os dados ainda não estão todos centralizados, portanto, teoricamente você ainda pode cometer um delito em cada cidade do brasil <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_wink.gif' alt=';)' class='wp-smiley' />  )</p>
<p>- um estudo da IDC, muito bem apresentado com conteúdo técnico e gerencial bem dosado falando sobre as próximas tendências do mercado, e ainda sugerindo métodos para você virtualizar seu ambiente com a maior segurança possível;</p>
<p>- Tecnologia Quad-Core da Intel, mostrando os aspectos voltados à virtualização e prometendo novidades enormes ainda este ano;</p>
<p>- Case da PRODAM, mostrando os desafios de migrar um Data Center de local, arruma-lo e ainda Virtualizar para abstrair o Hardware. Não posso deixar de citar a quantidade de equipamentos velhos e obsoletos que faziam parte do DC deles, e também as duas toneladas de cabos de rede desnecessários que foram removidos durante a migração, pois os comentários que se escutava era: &#8220;está vendo o que é feito com o seu dinheiro dos impostos?&#8221; <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_cool.gif' alt='8)' class='wp-smiley' /> </p>
<p>Acho ainda interessante dizer que na parte da tarde, a cada palestra de 1 hora o pessoal da vmware abriu um espaço de 15 minutos para algum parceiro deles falaram um pouco sobre a empresa, e por incrível que pareça, esses 15 minutos foram extremamente técnicos (NetApp, EMC e Sun com o Rafael Tinoco, que mandou muito bem no <a href="http://www.everlinux.com/blog/category/fisl/fisl90/">FISL 9.0</a>) alegrando eu e mais o restante de público Nerd não-gerente ali presente.</p>
<p>Claro que eu já esperava que o evento seria mais voltado à executivos de Terno e Gravata do que para Nerds de Jeans e Camiseta, portanto não posso dizer que estou decepcionado pois eu sabia que isso não era nenhum FISL. Muito pelo contrário, o lugar foi muito bem escolhido, a organização foi ótima e o evento bastante proveitoso.</p>
<p>Também serviu para encontrar velhos amigos (ex-professores do meu colégio, colegas de classe da faculdade, amigos de outras empresas e figuras que eu só encontro no FISL), além de perder um pouco o medo do inglês tentando fazer uns gringos entender que na maior parte das vezes, nós não usamos Linux porque é &#8220;de graça&#8221; ou é mais barato do que a solução X ou Y &#8212; muito pelo contrário, na maior parte das vezes nós pagamos pelo Linux (SuSE, Red Hat) mas usamos devido a suas características serem superiores do que as dos concorrentes, como por exemplo Desempenho, Estabilidade e Segurança.</p>
<p>E foi bom saber também que as empresas estão correndo atrás em adaptar seus produtos/ soluções para Linux nos próximos meses (o <a href="http://www.symmetry.com.br/vizioncore_vconverter.php">VConverter da VizionCore</a> me parece muito promissor). Além do mais, as empresas que já tem solução para Red Hat ainda não expandiram a mesma para SuSE ou Ubuntu devido a problemas gerenciais, como o treinamento do pessoal de suporte e etc, e não por problemas técnicos.</p>
<p>Muito interessante também o pessoal da NetAPP elogiando o iSCSI que geralmente não é bem vindo por ser uma alternativa barata às caríssimas soluções com FC/HBA&#8230;</p>
<p>E para fechar, uma feature do VmWare que eu gostei muito e sinto falta no Xen é o chamado &#8220;VmWare LifeCycle Manager&#8221; que irá cuidar que as suas máquinas virtuais não se proliferem e espalhem em seu ambiente que nem coelhos, pois como é muito fácil provisionar a mesma é também muito fácil você perder o controle e não ter mais certeza se a VM está ou não sendo utilizada ainda&#8230;</p>
<p>Bom, vou parando por aqui e peço desculpas se o texto ficou muito grande ou com as idéias desordenadas&#8230; agora preciso descansar pois amanhã terei mais uma migração grande para fazer de madrugada, portanto, um dia desses eu arrumo melhor este texto <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_wink.gif' alt=';)' class='wp-smiley' /> </p>
<p>Boa noite para vocês<br />
Diga &#8220;WI!!!!!!!&#8221; <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_biggrin.gif' alt=':D' class='wp-smiley' /><br />
PS: O console só chega semana que vem&#8230;</p>
]]></content:encoded>
			<wfw:commentRss>http://everlinux.com/blog/2008/06/19/vmware-virtualization-forum-2008-eu-fui/feed</wfw:commentRss>
		<slash:comments>6</slash:comments>
		</item>
		<item>
		<title>RedHat Enterprise Linux 5.1 Cluster Suite</title>
		<link>http://everlinux.com/blog/2008/04/22/redhat-enterprise-linux-51-cluster-suite</link>
		<comments>http://everlinux.com/blog/2008/04/22/redhat-enterprise-linux-51-cluster-suite#comments</comments>
		<pubDate>Tue, 22 Apr 2008 14:47:32 +0000</pubDate>
		<dc:creator>Tiago Cruz</dc:creator>
				<category><![CDATA[Cluster]]></category>
		<category><![CDATA[Red Hat]]></category>
		<category><![CDATA[Storage]]></category>

		<guid isPermaLink="false">http://www.everlinux.com/blog/2008/04/22/redhat-enterprise-linux-51-cluster-suite/</guid>
		<description><![CDATA[Por Fábio Silva em seu blog: &#8220;Recentemente precisei configurar um ambiente com RHEL 5.1 Cluster Suite. Em contato com o pessoal do canal #linux-cluster no irc.freenode.net, eles me esclareceram algumas dúvidas e então consegui montar o ambiente. Assim, resolvi criar um howto em português para que fosse disponibilizado na página do projeto de cluster da [...]]]></description>
			<content:encoded><![CDATA[<p>Por Fábio Silva em <a href="http://www.fabiosilva.eti.br/">seu blog</a>:</p>
<p>&#8220;Recentemente precisei configurar um ambiente com RHEL 5.1 Cluster Suite.</p>
<p>Em contato com o pessoal do canal #linux-cluster no irc.freenode.net, eles me esclareceram algumas dúvidas e então consegui montar o ambiente.</p>
<p>Assim, resolvi criar um howto em português para que fosse disponibilizado na página do projeto de cluster da redhat, e aqui está o link para os interessados.&#8221;</p>
<p><a href="http://sources.redhat.com/cluster/wiki/QuickStart-Portuguese">http://sources.redhat.com/cluster/wiki/QuickStart-Portuguese</a></p>
<p>Este post é somente para agregar mais uns links interessantes sobre o assunto:</p>
<p><a href="http://sources.redhat.com/cluster/wiki/FAQ/Fencing">FAQ Sobre Fencing Devices</a></p>
<p><a href="http://wiki.samba.org/index.php/6.0._DRBD">Replicated Failover Domain Controller and file server using LDAP</a></p>
<p><a href="https://open.datacore.ch/DCwiki.open/Wiki.jsp?page=GFS.GNBD.Usage">How to use GNBD to export and import devices for GFS</a></p>
<p><a href="http://http://www.redhatmagazine.com/2007/03/19/teaching-your-cluster-and-storage-systems-to-dance-an-introduction-to-conga/">Teaching your cluster and storage systems to dance: An introduction to Conga</a></p>
<p><a href="http://www.redhat.com/docs/manuals/csgfs/admin-guide/cs-gfs-cons.html">Using Red Hat GFS with Red Hat Cluster Suite</a></p>
<p>Por último, mas não menos importante, um &#8220;txt&#8221; tosco, mas nem por isso com menas informações do que os demais: <a href="http://people.redhat.com/teigland/cluster2-arch.txt">cluster2-arch.txt</a>. Lá existe uma seção muito esclarecedora sobre <strong>&#8220;Fencing: What, When, Why&#8221;</strong>.</p>
]]></content:encoded>
			<wfw:commentRss>http://everlinux.com/blog/2008/04/22/redhat-enterprise-linux-51-cluster-suite/feed</wfw:commentRss>
		<slash:comments>4</slash:comments>
		</item>
		<item>
		<title>Storage AMS500 da Hitachi</title>
		<link>http://everlinux.com/blog/2008/02/01/storage-ams500-da-hitachi</link>
		<comments>http://everlinux.com/blog/2008/02/01/storage-ams500-da-hitachi#comments</comments>
		<pubDate>Fri, 01 Feb 2008 19:27:30 +0000</pubDate>
		<dc:creator>Tiago Cruz</dc:creator>
				<category><![CDATA[Linux]]></category>
		<category><![CDATA[Storage]]></category>

		<guid isPermaLink="false">http://www.everlinux.com/blog/2008/02/01/storage-ams500-da-hitachi/</guid>
		<description><![CDATA[Este post abrange o Storage AMS500 da Hitachi, comercialmente conhecido pelo modelo DF700M. O Storage possui RAID Groups em Fibre Channel usando RAID5 (5D+1P), e algumas RAID Groups de discos SATA usando RAID6 (7D+2P). Sobre o Multi Path (HDLM) do Storage Hitachi Conforme o desenho, pode-se observar que será necessário um software multipath para controlar [...]]]></description>
			<content:encoded><![CDATA[<p>Este post abrange o <a href="http://www.hds.com/products/storage-systems/adaptable-modular-storage.html">Storage AMS500</a> da Hitachi, comercialmente conhecido pelo modelo DF700M.</p>
<p>O Storage possui RAID Groups em <a href="http://pt.wikipedia.org/wiki/Fibre_channel">Fibre Channe</a>l usando RAID5 (5D+1P), e algumas RAID Groups de discos <a href="http://pt.wikipedia.org/wiki/Serial_ATA">SATA</a> usando RAID6 (7D+2P).</p>
<p><a href='http://www.everlinux.com/blog/wp-content/uploads/2008/02/ams500.jpg' title='AMS 500'><img src='http://www.everlinux.com/blog/wp-content/uploads/2008/02/ams500.thumbnail.jpg' alt='AMS 500' /></a></p>
<p><strong>Sobre o Multi Path (HDLM) do Storage Hitachi</strong></p>
<p>Conforme o desenho, pode-se observar que será necessário um software multipath para controlar os vários caminhos que os servidores &#8220;server-spo-la-1&#8243; e &#8220;server-spo-la-2&#8243; podem fazer para chegar até os discos do storage.</p>
<p>Para exemplificar, o server-spo-la-1 pode fazer os seguintes caminhos:<br />
- Passar pela fibra saindo pela HBA0 até o Switch 1, e chegar no storage pela Controller 0, porta A;<br />
- Passar pela fibra saindo pela HBA1 até o Switch 2, e chegar no storage pela Controller 0, porta B;<br />
- Passar pela fibra saindo pela HBA1 até o Switch 2, e chegar no storage pela Controller 1, porta A;<br />
- Passar pela fibra saindo pela HBA0 até o Switch 1, e chegar no storage pela Controller 0, porta B;</p>
<p>É aí que entra o HDML (<em>Hitachi Dynamic Link Manager</em>). O problema é que para cada caminho que o SO faz até o Storage, ele encontra discos e &#8220;pensa&#8221; que são discos diferentes. Portanto, se você tem 25 LUNs exportadas para o server-spo-la-1, ele irá &#8220;enxergar&#8221; 100 discos (25*4=100). Será algo parecido com isso:</p>
<p><code>root@server-spo-la-1 ~# fdisk -l 2> /dev/null | grep "38.6 GB"<br />
Disk /dev/sdb: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdc: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdd: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sde: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdf: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdg: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdh: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdi: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdj: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdk: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdl: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdm: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdn: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdo: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdp: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdq: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdr: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sds: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdt: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdu: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdv: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdw: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdx: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdy: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdz: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdaa: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdab: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdac: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdad: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdae: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdaf: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sdag: 38.6 GB, 38654705664 bytes</code><br />
&#8230;</p>
<p>Completamente bizarro <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_cool.gif' alt='8-)' class='wp-smiley' /> </p>
<p><strong>Instalação do HDLM</strong></p>
<p>O processo de instalação é um tanto chato, e cada servidor precisa de uma licença diferente. Em termos gerais, seria mais ou menos isso:</p>
<p>-> Verificar se o server acha os discos com um fdisk -l</p>
<p>-> Copiar o license key para:<br />
/var/tmp/hdlm_license<br />
/etc/opt/DynamicLinkManager/dlm.lic_key</p>
<p>-> Rodar o /media/cdrom/installhdml</p>
<p>-> No /etc/profile</p>
<blockquote><p>PATH=$PATH:/opt/DynamicLinkManager/bin ; export PATH</p></blockquote>
<p>-> Reboot</p>
<p>-> Verificar se o fdisk -l mostra mais discos <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_smile.gif' alt=':)' class='wp-smiley' /> </p>
<p>Se tudo der certo, além dos 100 discos de outrora, seu servidor irá enxergar mais 25 (!!!) mas agora com outro nome. E será esse device que você irá utilizar para guardar os dados:</p>
<blockquote><p>Disk /dev/sddlmaa: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sddlmab: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sddlmac: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sddlmad: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sddlmae: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sddlmaf: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sddlmag: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sddlmah: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sddlmai: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sddlmaj: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sddlmak: 38.6 GB, 38654705664 bytes<br />
Disk /dev/sddlmal: 38.6 GB, 38654705664 bytes</p></blockquote>
<p>&#8230;</p>
<p><strong>Configuração do HDLM</strong></p>
<p>Caso você queira testar se o HDML está mesmo funcionando, você poderia tirar uma fibra enquanto copia os dados, chutar o switch ou algo do tipo <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_smile.gif' alt=':-)' class='wp-smiley' /> </p>
<blockquote><p># dlnkmgr set -lb on<br />
# dlnkmgr set -pchk on -intvl 5<br />
# dlnkmgr set -afb on -intvl 5<br />
# dlnkmgr set -ellv 2<br />
# dlnkmgr set -systflv 1<br />
# dlnkmgr set -elfs 1000<br />
# dlnkmgr set -elfn 5<br />
# dlnkmgr set -systfs 2000<br />
# dlnkmgr set -systfn 10
</p></blockquote>
<p>Para verificar:</p>
<blockquote><p>
# dlnkmgr view -path | more<br />
# dlnkmgr view -sys<br />
# dlnkmgr view -lu | more
</p></blockquote>
<p>Este setup irá ativar o Load Balance, verificação de PATH, auto Failback em caso de falhas e etc.</p>
<p><strong>Configurando o LVM</strong></p>
<p>Existe ainda mais um cuidado a ser tomado antes de você conseguir brincar com o Storage. O LVM é uma ferramenta muito esperta, e ele identifica cada disco com uma assinatura, individual e intransferível, mais ou menos como a sua impressão digital.</p>
<p>Portanto, ao adicionar um disco do HDLM ao LVM, você verá algo parecido com isso:</p>
<blockquote><p>root@server-la-2 ~# pvcreate /dev/sddlmaa<br />
Physical volume &#8220;/dev/sddlmaa&#8221; successfully created</p>
<p>root@server-la-2 ~# pvs<br />
Found duplicate PV sXzKX9RMVxSafKzfDJP78Zr4qYFPNPcO: using /dev/sdb not /dev/sddlmaa<br />
Found duplicate PV sXzKX9RMVxSafKzfDJP78Zr4qYFPNPcO: using /dev/sdch not /dev/sdb<br />
Found duplicate PV sXzKX9RMVxSafKzfDJP78Zr4qYFPNPcO: using /dev/sdbf not /dev/sdch<br />
Found duplicate PV sXzKX9RMVxSafKzfDJP78Zr4qYFPNPcO: using /dev/sdad not /dev/sdbf<br />
Found duplicate PV sXzKX9RMVxSafKzfDJP78Zr4qYFPNPcO: using /dev/sddlmaa not /dev/sdad<br />
PV VG Fmt Attr PSize PFree<br />
/dev/sda2 Vol_LVM lvm2 a- 408.17G 0<br />
/dev/sddlmaa lvm2 — 36.00G 36.00G
</p></blockquote>
<p>O que acontece é que o LVM está vendo o mesmo disco (mesma LUN, na verdade) pelos 04 caminhos diferentes e mais o novo caminho, do HDLM.</p>
<p>O que você deve fazer, é editar o arquivo /etc/lvm/lvm.conf e criar um filtro separando os devices que devem de fato serem manipulados pelo LVM, algo deste tipo:</p>
<blockquote><p>filter = [ "a/sda1-9$/" "a/sddla-za-za-z$/" "r/.*/" ]</p></blockquote>
<p>Após isso, basta rodar um &#8216;vgscan -vv&#8217; para refazer o cache.</p>
<p><strong>Passos rápidos para configurar o LVM no Storage</strong></p>
<p>Papo rápido:</p>
<p><code><br />
# for i in `fdisk -l 2> /dev/null | grep "sddl" | awk '{print $2}' | cut -d : -f 1`; do pvcreate $i; done<br />
</code></p>
<p><code><br />
# vgcreate -Ay AMS500 /dev/sddlmaa /dev/sddlmab /dev/sddlmac /dev/sddlmad \<br />
/dev/sddlmae /dev/sddlmaf /dev/sddlmag /dev/sddlmah /dev/sddlmai /dev/sddlmaj \<br />
/dev/sddlmak /dev/sddlmal /dev/sddlmam /dev/sddlman /dev/sddlmao /dev/sddlmap \<br />
/dev/sddlmba /dev/sddlmbb /dev/sddlmbc /dev/sddlmbd /dev/sddlmbe /dev/sddlmbf \<br />
/dev/sddlmbg /dev/sddlmbh /dev/sddlmbi /dev/sddlmbj /dev/sddlmbk /dev/sddlmbl<br />
</code></p>
<p><code><br />
# lvcreate -Ay -L 1.2T --name u01 AMS500</p>
<p># mkfs.ext3 /dev/AMS500/u01</p>
<p># mkdir /u01; mount /dev/AMS500/u01 /u01</code></p>
<p>Pronto! 1.2 TB para brincar <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_biggrin.gif' alt=':-D' class='wp-smiley' /> </p>
<p><strong>Ganhando desempenho fazendo Stripe no LVM</strong></p>
<p>É possível fazer Stripe do lado do Sistema Operacional, visando um maior desempenho <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_smile.gif' alt=':)' class='wp-smiley' /> </p>
<p>Cada servidor está &#8220;enxergando&#8221; 28 LUNs do Storage, como se cada LUN fosse um disco.</p>
<p>Utilizei o &#8220;dd&#8221; para criar arquivos de 10 e 40 GB nos servidores, tanto na &#8220;Barriga&#8221; da máquina (Dell <a href="http://www.dell.com/content/products/productdetails.aspx/pedge_2950">PowerEdge 2950</a> em RAID 10) quanto no Storage. Depois disso, irei utilizar nos servidores Stripes com granulação de 128 KB, sempre no FileSystem com Block Size de 4KB para comparar:</p>
<p><em># Teste Storage Billing<br />
# LVM padrão (sem stripe) e Block Size de 4KB</em></p>
<p>=> 10 GB com LVM default:<br />
root@server-la-1 ~# Storage: 1m38.939s<br />
root@server-la-1 ~# Barriga: 1m22.395s</p>
<p>root@server-la-2 ~# Storage: 1m57.203s<br />
root@server-la-2 ~# Barriga: 1m15.134s</p>
<p>=> 40 GB com LVM default:<br />
root@server-la-1 ~# Storage: 10m23.250s<br />
root@server-la-1 ~# Barriga: 06m04.812s</p>
<p>root@server-la-2 ~# Storage: 11m22.150s<br />
root@server-la-2 ~# Barriga: 06m01.234s</p>
<p><em># lvcreate -Ay -i 28 -I 128 -l258020 &#8211;name u01 AMS500<br />
# 28 Stripes com granulação de 128 KB cada (Block size=4096)</em></p>
<p>=> 10 GB com LVM + Stripe 128K<br />
root@server-la-2 ~# Storage: 1m02.411s<br />
root@server-la-2 ~# Barriga: 1m18.440s</p>
<p>=> 40 GB com LVM + Stripe 128K<br />
root@server-la-2 ~# Storage: 4m49.578s<br />
root@server-la-2 ~# Barriga: 5m50.452s</p>
<p><em># lvcreate -Ay -i 28 -I 64 -l258020 &#8211;name u01 AMS500<br />
# 28 Stripes com granulação de 64 KB cada (Block size=4096)</em></p>
<p>=> 10 GB com LVM + Stripe 64K<br />
root@server-la-1 ~# Storage: 1m42.396s<br />
root@server-la-1 ~# Barriga: 1m44.903s</p>
<p>=> 40 GB com LVM + Stripe 64K<br />
root@server-la-1 ~# Storage: 5m04.977s<br />
root@server-la-1 ~# Barriga: 5m55.737s</p>
<p><strong>Teste de desempenho</strong></p>
<p>Resultados dos testes de desempenho do Storage feitos por uma ferramenta mais profissional (Bonnie++), que faz vários testes além do qual eu fiz:</p>
<p><a href="http://everlinux.com/raid/bonnie.html">http://everlinux.com/raid/bonnie.html</a></p>
<p>Testei na minha máquina de casa (IDE) e do escritório (SATA) também, só para lembrar que eu preciso trocar meu velho Athlon 2000 <img src='http://everlinux.com/blog/wp-includes/images/smilies/icon_smile.gif' alt=':-)' class='wp-smiley' /> </p>
]]></content:encoded>
			<wfw:commentRss>http://everlinux.com/blog/2008/02/01/storage-ams500-da-hitachi/feed</wfw:commentRss>
		<slash:comments>6</slash:comments>
		</item>
	</channel>
</rss>
<!-- This Quick Cache file was built for (  everlinux.com/blog/category/storage/feed ) in 1.10544 seconds, on Jul 30th, 2010 at 8:46 pm UTC. -->
<!-- This Quick Cache file will automatically expire ( and be re-built automatically ) on Jul 30th, 2010 at 9:46 pm UTC -->