doc/xml/QuickStartUnix/appendix_dafs.xml

   1 <?xml version="1.0" encoding="UTF-8"?>
   2 <appendix id="DAFS">
   3   <title>The Demand-Attach File Server</title>
   4
   5   <para>This chapter explains and describes the Demand-Attach File Server (DAFS
   6   for short), a new feature in OpenAFS 1.6. A few things are different when
   7   setting up and maintaining a DAFS fileserver vs a traditional fileserver, so
   8   it is important to know which one you are running.</para>
   9
  10   <para>If you are trying to decide whether or not to run DAFS on a fileserver,
  11   note that you can switch between DAFS and the traditional fileserver fairly
  12   easily at any time. Aside from performance differences, the two fileservers
  13   generally look and act similarly, so choosing one over the other in the
  14   beginning should not hurt you later on. So, if you are not overly concerned
  15   with performance, you can just pick one and follow the directions for it, and
  16   you should be fine.</para>
  17
  18   <sect1 id="DAFS001">
  19     <title>Justification and Background</title>
  20
  21     <para>DAFS changes many things with how the fileserver and other server
  22     processes access volumes on disk. Most of these changes are only of concern
  23     to developers, as there are no visible differences to users or
  24     administrators. A few changes are only of concern to administrators while
  25     debugging a problem, and only one or two changes affect the day-to-day
  26     administration of a fileserver. See the sections on <link
  27     linkend="DAFS002">DAFS Binaries</link> and <link
  28     linkend="DAFS003">Salvaging</link> for the main functional
  29     differences.</para>
  30
  31     <para>Among other things, DAFS provides a new way for the fileserver and
  32     other server programs to access volumes. It does not change how file data is
  33     accessed inside a volume (like namei or inode do), but rather how a volume
  34     as a whole is accessed by the fileserver. When a traditional fileserver is
  35     started, it will locate and attach all AFS volumes it can find on all AFS
  36     partitions on the server. This attachment process involves reading the
  37     volume header and setting a flag to say that the volume is in use. On a
  38     clean shutdown, the fileserver detach all volumes it attached by clearing
  39     that flag. If the fileserver encounters a volume where the flag is already
  40     set before it attached the volume, it will know that volume was not detached
  41     cleanly, and that it needs to be salvaged.</para>
  42
  43     <para>This process of attaching and detaching at startup and shutdown takes
  44     time. When fileservers start to have thousands of volumes, it can take
  45     several minutes to start or cleanly shutdown. Additionally, when a
  46     fileserver is shutdown uncleanly, all of the volumes on the server must be
  47     salvaged before the fileserver can be brought up again. Even volumes that
  48     nobody has ever accessed must be salvaged.</para>
  49
  50     <para>To improve this situation, DAFS does not attach all volumes on
  51     startup. Instead, volumes are only attached when they are first accessed,
  52     and are detached if idle for a long period of time (see the -vlru* options
  53     for the fileserver; this is configurable but is generally several
  54     hours).</para>
  55
  56     <para>Thus, after an unclean shutdown with DAFS, only the volumes in use at
  57     the time of the shutdown need to be salvaged. In addition, the fileserver
  58     does not need to wait for salvaging to finish before starting up. Instead of
  59     salvaging everything before the fileserver starts, the fileserver causes
  60     salvages to be issued on a volume when a damanged volume is accessed (called
  61     "demand-salvages").</para>
  62
  63     <para>The result of all of this is that a fileserver running DAFS can
  64     generally be started and stopped in a manner of seconds, where a traditional
  65     fileserver could take dozens of minutes, or even hours in the case of an
  66     unclean shutdown.</para>
  67   </sect1>
  68   <sect1 id="DAFS002">
  69     <title>DAFS Binaries</title>
  70
  71     <para>OpenAFS ships with binaries for DAFS and non-DAFS alongside each
  72     other. Programs that exist in both DAFS and non-DAFS variants typically have
  73     a <emphasis role="bold">da</emphasis> prefix to indicate the DAFS variant.
  74     For example, <emphasis role="bold">dafileserver</emphasis> is the DAFS
  75     version of the traditional <emphasis role="bold">fileserver</emphasis>
  76     binary. Similarly for <emphasis role="bold">davolserver</emphasis> and
  77     <emphasis role="bold">volserver</emphasis>, <emphasis
  78     role="bold">dasalvager</emphasis> and <emphasis
  79     role="bold">salvager</emphasis>, and even some other tools like <emphasis
  80     role="bold">dafssync-debug</emphasis> and <emphasis
  81     role="bold">fssync-debug</emphasis>.</para>
  82
  83     <para>A DAFS-enabled fileserver will run the <emphasis
  84     role="bold">dafs</emphasis> bnode in bosserver, instead of the traditional
  85     fileserver's <emphasis role="bold">fs</emphasis> bnode. The <emphasis
  86     role="bold">dafs</emphasis> bnode runs four different binaries as opposed to
  87     the <emphasis role="bold">fs</emphasis> bnode's three. Three of the programs
  88     in the <emphasis role="bold">dafs</emphasis> bnode are just DAFS equivalents
  89     of the corresponding <emphasis role="bold">fs</emphasis> bnode programs, but
  90     the fourth one is an entirely new daemon that has no non-DAFS equivalent:
  91     the salvageserver.</para>
  92
  93     <informaltable>
  94       <tgroup cols="2">
  95         <colspec colwidth="50*" />
  96         <colspec colwidth="50*" />
  97         <thead>
  98           <row>
  99             <entry><emphasis role="bold">Traditional binary</emphasis></entry>
 100             <entry><emphasis role="bold">DAFS binary</emphasis></entry>
 101           </row>
 102         </thead>
 103
 104         <tbody>
 105           <row>
 106             <entry><emphasis role="bold">/usr/afs/bin/fileserver</emphasis></entry>
 107             <entry><emphasis role="bold">/usr/afs/bin/dafileserver</emphasis></entry>
 108           </row>
 109           <row>
 110             <entry><emphasis role="bold">/usr/afs/bin/volserver</emphasis></entry>
 111             <entry><emphasis role="bold">/usr/afs/bin/davolserver</emphasis></entry>
 112           </row>
 113           <row>
 114             <entry>No equivalent</entry>
 115             <entry><emphasis role="bold">/usr/afs/bin/salvageserver</emphasis></entry>
 116           </row>
 117           <row>
 118             <entry><emphasis role="bold">/usr/afs/bin/salvager</emphasis></entry>
 119             <entry><emphasis role="bold">/usr/afs/bin/dasalvager</emphasis></entry>
 120           </row>
 121         </tbody>
 122       </tgroup>
 123     </informaltable>
 124   </sect1>
 125
 126   <sect1 id="DAFS003">
 127     <title>Salvaging</title>
 128
 129     <para>With a traditional fileserver, salvages usually occur in two
 130     situations:
 131       <orderedlist>
 132         <listitem>
 133           <para>The fileserver shuts down uncleanly, and when brought back up,
 134           all partitions are salvaged before the fileserver is available.</para>
 135         </listitem>
 136         <listitem>
 137           <para>A volume experiences some corruption after it has been brought
 138           online, and an administrator manually schedules a salvage for an
 139           individual volume with <emphasis role="bold">bos salvage</emphasis>.
 140           Usually the way you notice this is that the fileserver noticed a
 141           volume has become corrupt and has taken it offline.</para>
 142         </listitem>
 143       </orderedlist>
 144     </para>
 145
 146     <para>With DAFS, neither of these occur in normal operation. With DAFS, a
 147     volume will be salvaged automatically when a problem is detected in a
 148     volume. This occurs whether the fileserver detects the volume was not
 149     detached cleanly, or if the fileserver detects corrupt volume data.</para>
 150
 151     <para>In normal DAFS operation, you should not need to ever run <emphasis
 152     role="bold">bos salvage</emphasis>. However, if you suspect a bug, or that
 153     there is corruption in a volume that the fileserver has not detected, you
 154     can run <emphasis role="bold">bos salvage</emphasis> to manually issue a
 155     salvage.</para>
 156
 157     <para>All salvages scheduled from the salvageserver will be logged in
 158     <emphasis role="bold">/usr/afs/logs/SalsrvLog</emphasis>, in the same format
 159     as salvages for traditional fileservers are logged. If you issue a
 160     whole-partition or whole-server salvage, the logs for that will still be
 161     located in <emphasis role="bold">/usr/afs/logs/SalvageLog</emphasis>.</para>
 162   </sect1>
 163
 164   <sect1 id="DAFS004">
 165     <title>Converting a Fileserver to DAFS</title>
 166
 167     <para>If you have an existing traditional fileserver, you can convert it to
 168     DAFS fairly easily. The conversion is in-place, but there is at least a
 169     small amount of downtime involved even if nothing goes wrong, since the
 170     fileserver processes must be restarted. If you want to avoid any downtime,
 171     move any volumes on the fileserver to another fileserver before performing
 172     the conversion.</para>
 173
 174     <orderedlist>
 175       <listitem>
 176         <para>If the fileserver is currently running a version of OpenAFS
 177         earlier than 1.6.0, upgrade the fileserver to a version in the 1.6
 178         series or later. This is not strictly necessary, but if you encounter
 179         problems here, it is helpful to catch them before converting to DAFS so
 180         as to isolate the problem.</para>
 181
 182         <para>If you do not upgrade the fileserver before converting to DAFS,
 183         install the new OpenAFS binaries somewhere on the server.</para>
 184       </listitem>
 185
 186       <listitem>
 187         <para>If the current bosserver process running on the fileserver is
 188         older than OpenAFS 1.6.0, you must upgrade it first. This can be done by
 189         replacing the bosserver binary and running <emphasis role="bold">bos
 190         restart -bosserver</emphasis>.  This will cause <emphasis>all</emphasis>
 191         AFS processes controlled by the bosserver to be restarted. You can do
 192         this at the same as upgrading the fileserver in the previous step if
 193         desired, to avoid restarting the fileserver twice.</para>
 194       </listitem>
 195
 196       <listitem>
 197         <para>Shutdown and stop the old <emphasis role="bold">fs</emphasis> bnode:
 198           <programlisting>
 199  # <emphasis role="bold">bos stop </emphasis>&lt;<replaceable>machine name</replaceable>&gt; <emphasis role="bold">fs -wait -cell </emphasis>&lt;<replaceable>cell name</replaceable>&gt;
 200           </programlisting>
 201         </para>
 202       </listitem>
 203
 204       <listitem>
 205         <para>Create and start the new <emphasis role="bold">dafs</emphasis> bnode.
 206           <programlisting>
 207  # <emphasis role="bold">bos create  </emphasis>&lt;<replaceable>machine name</replaceable>&gt; <emphasis role="bold">dafs dafs</emphasis> \
 208                <emphasis role="bold">-cmd</emphasis> '<emphasis role="bold">/usr/afs/bin/dafileserver</emphasis> &lt;<replaceable>dafileserver options</replaceable>&gt;' \
 209                <emphasis role="bold">-cmd</emphasis> '<emphasis role="bold">/usr/afs/bin/davolserver</emphasis> &lt;<replaceable>davolserver options</replaceable>&gt;' \
 210                <emphasis role="bold">-cmd</emphasis> '<emphasis role="bold">/usr/afs/bin/salvageserver</emphasis> &lt;<replaceable>salvageserver options</replaceable>&gt;' \
 211                <emphasis role="bold">-cmd</emphasis> '<emphasis role="bold">/usr/afs/bin/dasalvager</emphasis> &lt;<replaceable>dasalvager options</replaceable>&gt;' \
 212                <emphasis role="bold">-cell</emphasis> &lt;<replaceable>cell name</replaceable>&gt;
 213           </programlisting>
 214         You can use the same options for the <emphasis
 215         role="bold">dafileserver</emphasis> process as you did for the <emphasis
 216         role="bold">fileserver</emphasis> process, and the same options for
 217         <emphasis role="bold">davolserver</emphasis> as you did for <emphasis
 218         role="bold">volserver</emphasis>.</para>
 219
 220         <para>You can also use most of the same options for the <emphasis
 221         role="bold">salvageserver</emphasis> and <emphasis
 222         role="bold">dasalvager</emphasis> processes as you did for the <emphasis
 223         role="bold">salvager</emphasis> process; see their respective man pages.
 224         However, if you are upgrading from the 1.4.x series of OpenAFS, be aware
 225         that the <emphasis role="bold">-DontSalvage</emphasis> option does not
 226         exist anymore for the salvager (with or without DAFS).</para>
 227
 228         <para>Also note that the <emphasis role="bold">dafileserver</emphasis>
 229         process has many more options to tune dafs-specific parameters. See the
 230         <emphasis role="bold">dafileserver</emphasis> man page for information
 231         about them.</para>
 232       </listitem>
 233     </orderedlist>
 234
 235     <para>After you have performed these steps, switching back and forth between
 236     running a DAFS and a traditional fileserver is as simple a running <emphasis
 237     role="bold">bos stop</emphasis> on one bnode, and <emphasis role="bold">bos
 238     start</emphasis> on the other. Once you are confident that the DAFS
 239     processes are running smoothly and you do not anticipate switching back to
 240     the traditional fileserver, you can <emphasis role="bold">bos
 241     delete</emphasis> the <emphasis role="bold">fs</emphasis> bnode.</para>
 242   </sect1>
 243 </appendix>