/usr/share/doc/slony1-2-doc/adminguide/failover.html

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN""http://www.w3.org/TR/html4/loose.dtd">
<HTML
><HEAD
><TITLE
>Doing switchover and failover with Slony-I</TITLE
><META
NAME="GENERATOR"
CONTENT="Modular DocBook HTML Stylesheet Version 1.79"><LINK
REV="MADE"
HREF="mailto:slony1-general@lists.slony.info"><LINK
REL="HOME"
TITLE="Slony-I 2.2.5 Documentation"
HREF="index.html"><LINK
REL="UP"
TITLE="Administration Tasks"
HREF="administration.html"><LINK
REL="PREVIOUS"
TITLE="Database Schema Changes (DDL)"
HREF="ddlchanges.html"><LINK
REL="NEXT"
TITLE="Advanced Concepts"
HREF="advanced.html"><LINK
REL="STYLESHEET"
TYPE="text/css"
HREF="stylesheet.css"><META
HTTP-EQUIV="Content-Type"
CONTENT="text/html; charset=ISO-8859-1"></HEAD
><BODY
CLASS="SECT1"
><DIV
CLASS="NAVHEADER"
><TABLE
SUMMARY="Header navigation table"
WIDTH="100%"
BORDER="0"
CELLPADDING="0"
CELLSPACING="0"
><TR
><TH
COLSPAN="5"
ALIGN="center"
VALIGN="bottom"
><SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
> 2.2.5 Documentation</TH
></TR
><TR
><TD
WIDTH="10%"
ALIGN="left"
VALIGN="top"
><A
HREF="ddlchanges.html"
ACCESSKEY="P"
>Prev</A
></TD
><TD
WIDTH="10%"
ALIGN="left"
VALIGN="top"
><A
HREF="administration.html"
>Fast Backward</A
></TD
><TD
WIDTH="60%"
ALIGN="center"
VALIGN="bottom"
>Chapter 3. Administration Tasks</TD
><TD
WIDTH="10%"
ALIGN="right"
VALIGN="top"
><A
HREF="administration.html"
>Fast Forward</A
></TD
><TD
WIDTH="10%"
ALIGN="right"
VALIGN="top"
><A
HREF="advanced.html"
ACCESSKEY="N"
>Next</A
></TD
></TR
></TABLE
><HR
ALIGN="LEFT"
WIDTH="100%"></DIV
><DIV
CLASS="SECT1"
><H1
CLASS="SECT1"
><A
NAME="FAILOVER"
>3.4. Doing switchover and failover with <SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
></A
></H1
><A
NAME="AEN1223"
></A
><DIV
CLASS="SECT2"
><H2
CLASS="SECT2"
><A
NAME="AEN1226"
>3.4.1. Foreword</A
></H2
><P
><SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
> is an asynchronous replication system.  Because of
that, it is almost certain that at the moment the current origin of a
set fails, the final transactions committed at the origin will have
not yet propagated to the subscribers.  Systems are particularly
likely to fail under heavy load; that is one of the corollaries of
Murphy's Law.  Therefore the principal goal is to
<SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>prevent</I
></SPAN
> the main server from failing.  The best
way to do that is frequent maintenance.</P
><P
> Opening the case of a running server is not exactly what we
should consider a <SPAN
CLASS="QUOTE"
>"professional"</SPAN
> way to do system
maintenance.  And interestingly, those users who found it valuable to
use replication for backup and failover purposes are the very ones
that have the lowest tolerance for terms like <SPAN
CLASS="QUOTE"
>"system
downtime."</SPAN
> To help support these requirements, <SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
> not
only offers failover capabilities, but also the notion of controlled
origin transfer.</P
><P
> It is assumed in this document that the reader is familiar with
the <A
HREF="slonik.html"
><SPAN
CLASS="APPLICATION"
>slonik</SPAN
></A
> utility and knows at least how to set up a
simple 2 node replication system with <SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
>.</P
></DIV
><DIV
CLASS="SECT2"
><H2
CLASS="SECT2"
><A
NAME="AEN1238"
>3.4.2. Controlled Switchover</A
></H2
><A
NAME="AEN1240"
></A
><P
> We assume a current <SPAN
CLASS="QUOTE"
>"origin"</SPAN
> as node1 with one
<SPAN
CLASS="QUOTE"
>"subscriber"</SPAN
> as node2 (<SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>e.g.</I
></SPAN
> -
slave).  A web application on a third server is accessing the database
on node1.  Both databases are up and running and replication is more
or less in sync.  We do controlled switchover using <A
HREF="stmtmoveset.html"
>SLONIK MOVE SET</A
>.</P
><P
></P
><UL
><LI
><P
> At the time of this writing switchover to another
server requires the application to reconnect to the new database.  So
in order to avoid any complications, we simply shut down the web
server.  Users who use <SPAN
CLASS="APPLICATION"
>pg_pool</SPAN
> for the
applications database connections merely have to shut down the
pool.</P
><P
> What needs to be done, here, is highly dependent on the way
that the application(s) that use the database are configured.  The
general point is thus: Applications that were connected to the old
database must drop those connections and establish new connections to
the database that has been promoted to the <SPAN
CLASS="QUOTE"
>"master"</SPAN
> role.  There
are a number of ways that this may be configured, and therefore, a
number of possible methods for accomplishing the change:</P
><P
></P
><UL
><LI
><P
> The application may store the name of the database in
a file.</P
><P
> In that case, the reconfiguration may require changing the
value in the file, and stopping and restarting the application to get
it to point to the new location.</P
></LI
><LI
><P
> A clever usage of DNS might involve creating a CNAME
<A
HREF="http://www.iana.org/assignments/dns-parameters"
TARGET="_top"
> DNS
record </A
> that establishes a name for the application to use to
reference the node that is in the <SPAN
CLASS="QUOTE"
>"master"</SPAN
> role.</P
><P
> In that case, reconfiguration would require changing the CNAME
to point to the new server, and possibly restarting the application to
refresh database connections.</P
></LI
><LI
><P
> If you are using <SPAN
CLASS="APPLICATION"
>pg_pool</SPAN
> or some
similar <SPAN
CLASS="QUOTE"
>"connection pool manager,"</SPAN
> then the reconfiguration
involves reconfiguring this management tool, but is otherwise similar
to the DNS/CNAME example above.  </P
></LI
></UL
><P
> Whether or not the application that accesses the database needs
to be restarted depends on how it is coded to cope with failed
database connections; if, after encountering an error it tries
re-opening them, then there may be no need to restart it. </P
></LI
><LI
><P
> A small <A
HREF="slonik.html"
><SPAN
CLASS="APPLICATION"
>slonik</SPAN
></A
> script executes the
following commands:

</P><PRE
CLASS="PROGRAMLISTING"
>lock set (id = 1, origin = 1);
wait for event (origin = 1, confirmed = 2, wait on=1);
move set (id = 1, old origin = 1, new origin = 2);
wait for event (origin = 1, confirmed = 2, wait on=1);</PRE
><P></P
><P
> After these commands, the origin (master role) of data set 1
has been transferred to node2.  And it is not simply transferred; it
is done in a fashion such that node1 becomes a fully synchronized
subscriber, actively replicating the set.  So the two nodes have
switched roles completely.</P
></LI
><LI
><P
> After reconfiguring the web application (or
<SPAN
CLASS="APPLICATION"
><A
HREF="help.html#PGPOOL"
> pgpool </A
></SPAN
>) to
connect to the database on node2, the web server is restarted and
resumes normal operation.</P
><P
> Done in one shell script, that does the application shutdown,
<SPAN
CLASS="APPLICATION"
>slonik</SPAN
>, move config files and startup all
together, this entire procedure is likely to take less than 10
seconds.</P
></LI
></UL
><P
> You may now simply shutdown the server hosting node1 and do
whatever is required to maintain the server.  When <A
HREF="slon.html"
><SPAN
CLASS="APPLICATION"
>slon</SPAN
></A
> node1 is restarted later, it will start replicating
again, and soon catch up.  At this point the procedure to switch
origins is executed again to restore the original
configuration.</P
><P
> This is the preferred way to handle things; it runs quickly,
under control of the administrators, and there is no need for there to
be any loss of data.</P
><P
> After performing the configuration change, you should, run the 
<A
HREF="deploymentconcerns.html#TESTSLONYSTATE"
>Section 5.1.1</A
> scripts in order to
validate that the cluster state remains in good order after this
change. </P
></DIV
><DIV
CLASS="SECT2"
><H2
CLASS="SECT2"
><A
NAME="AEN1283"
>3.4.3. Failover</A
></H2
><A
NAME="AEN1285"
></A
><P
> If some more serious problem occurs on the
<SPAN
CLASS="QUOTE"
>"origin"</SPAN
> server, it may be necessary to <A
HREF="stmtfailover.html"
>SLONIK FAILOVER</A
> to a backup server.  This is a highly
undesirable circumstance, as transactions <SPAN
CLASS="QUOTE"
>"committed"</SPAN
> on
the origin, but not applied to the subscribers, will be lost.  You may
have reported these transactions as <SPAN
CLASS="QUOTE"
>"successful"</SPAN
> to
outside users.  As a result, failover should be considered a
<SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>last resort</I
></SPAN
>.  If the <SPAN
CLASS="QUOTE"
>"injured"</SPAN
>
origin server can be brought up to the point where it can limp along
long enough to do a controlled switchover, that is
<SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>greatly</I
></SPAN
> preferable.</P
><P
> <SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
> does not provide any automatic detection for failed
systems.  Abandoning committed transactions is a business decision
that cannot be made by a database system.  If someone wants to put the
commands below into a script executed automatically from the network
monitoring system, well ... it's <SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>your</I
></SPAN
> data, and
it's <SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>your</I
></SPAN
> failover policy. </P
><P
></P
><UL
><LI
><P
>The <A
HREF="slonik.html"
><SPAN
CLASS="APPLICATION"
>slonik</SPAN
></A
> command
</P><PRE
CLASS="PROGRAMLISTING"
>failover (id = 1, backup node = 2);</PRE
><P></P
><P
> causes node2 to assume the ownership (origin) of all sets that
have node1 as their current origin.  If there should happen to be
additional nodes in the <SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
> cluster, all direct subscribers of
node1 are instructed that this is happening.
<SPAN
CLASS="APPLICATION"
>Slonik</SPAN
> will also query all direct
subscribers in order to determine out which node has the highest
replication status (<SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>e.g.</I
></SPAN
> - the latest committed
transaction) for each set, and the configuration will be changed in a
way that node2 first applies those final before actually allowing
write access to the tables.</P
><P
> In addition, all nodes that subscribed directly to node1 will
now use node2 as data provider for the set.  This means that after the
failover command succeeded, no node in the entire replication setup
will receive anything from node1 any more.</P
><DIV
CLASS="NOTE"
><BLOCKQUOTE
CLASS="NOTE"
><P
><B
>Note: </B
> Note that in order for node 2 to be considered as a
candidate for failover, it must have been set up with the <A
HREF="stmtsubscribeset.html"
>SLONIK SUBSCRIBE SET</A
> option <TT
CLASS="COMMAND"
>forwarding =
yes</TT
>, which has the effect that replication log data is
collected in <A
HREF="table.sl-log-1.html"
>sl_log_1</A
>/<A
HREF="table.sl-log-2.html"
>sl_log_2</A
> on node 2.  If replication log data is
<SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>not</I
></SPAN
> being collected, then failover to that node
is not possible. </P
><P
>Only nodes listed in the failover_candidates view
can be used as the failover target for an origin. This is discussed in further detail below.</P
></BLOCKQUOTE
></DIV
></LI
><LI
><P
> Reconfigure and restart the application (or
<SPAN
CLASS="APPLICATION"
>pgpool</SPAN
>) to cause it to reconnect to
node2.</P
></LI
><LI
><P
> Purge out the abandoned node </P
><P
> You will find, after the failover, that there are still a full
set of references to node 1 in <A
HREF="table.sl-node.html"
>sl_node</A
>, as well
as in referring tables such as <A
HREF="table.sl-confirm.html"
>sl_confirm</A
>;
since data in <A
HREF="table.sl-log-1.html"
>sl_log_1</A
>/<A
HREF="table.sl-log-2.html"
>sl_log_2</A
> is still present, <SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
> cannot
immediately purge out the node. </P
><P
> After the failover is complete and all nodes have been
reconfigured you can remove all remnants of node1's
configuration information with the <A
HREF="stmtdropnode.html"
>SLONIK DROP NODE</A
>
command:

</P><PRE
CLASS="PROGRAMLISTING"
>drop node (id = 1, event node = 2);</PRE
><P></P
><P
> Supposing the failure resulted from some catastrophic failure
of the hardware supporting node 1, there might be no
<SPAN
CLASS="QUOTE"
>"remains"</SPAN
> left to look at.  If the failure was not
<SPAN
CLASS="QUOTE"
>"total"</SPAN
>, as might be the case if the node had to be
abandoned due to a network communications failure, you will find that
node 1 still <SPAN
CLASS="QUOTE"
>"imagines"</SPAN
> itself to be as it was before the
failure.  See <A
HREF="failover.html#REBUILDNODE1"
>Section 3.4.7</A
> for more details on the
implications.</P
></LI
><LI
><P
> After performing the configuration change, you
should, as  run the <A
HREF="deploymentconcerns.html#TESTSLONYSTATE"
>Section 5.1.1</A
>
scripts in order to validate that the cluster state remains in good
order after this change. </P
></LI
></UL
></DIV
><DIV
CLASS="SECT2"
><H2
CLASS="SECT2"
><A
NAME="FAILOVERTARGETS"
>3.4.4. Failover Targets</A
></H2
><P
>An origin node can only be failed over to other nodes in the cluster that
are valid failover targets.  The failover targets for a node must meet the 
following conditions</P
><P
></P
><UL
><LI
><P
>The failover target must be a direct subscriber of all sets that
the failed node is an origin for</P
></LI
><LI
><P
>The failover target must have paths to all nodes that the
failed node has paths to</P
></LI
></UL
><P
>The view sl_failover_targets displays the valid failover targets for 
each origin node.  Clusters that have more than two nodes and would like
to have the option of using failover need to be setup in such a way that
valid failover targets exist for the various failure scenarios that 
they wish to support.</P
></DIV
><DIV
CLASS="SECT2"
><H2
CLASS="SECT2"
><A
NAME="MULTIPLEFAILURES"
>3.4.5. Multiple Node Failures</A
></H2
><P
>If multiple nodes fail at the same time, maybe because an entire data-center has failed,
then all failed nodes should be passed to the failover command. If we consider a cluster where  node 1 the origin of a set and provides a subscription to
node 2 and node 3 then node 2 provides a subscription to node 4, what should happen if  both nodes 1 and 2 fail?
Slony can be told about the failed nodes with the following command
</P><PRE
CLASS="PROGRAMLISTING"
>FAILOVER (node=(id=1, backup node=3), node=(id=2, backup node=3));</PRE
><P>

This command requires that a paths exist between node 3 and 4. It will then redirect node 4 to receive the subscription from node 3.
&#13;</P
></DIV
><DIV
CLASS="SECT2"
><H2
CLASS="SECT2"
><A
NAME="AEN1352"
>3.4.6. Automating <TT
CLASS="COMMAND"
> FAIL OVER </TT
></A
></H2
><A
NAME="AEN1355"
></A
><P
> If you do choose to automate <TT
CLASS="COMMAND"
>FAIL OVER </TT
>, it
is important to do so <SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>carefully.</I
></SPAN
> You need to have
good assurance that the failed node is well and truly failed, and you
need to be able to assure that the failed node will not accidentally
return into service, thereby allowing there to be two nodes out there
able to respond in a <SPAN
CLASS="QUOTE"
>"master"</SPAN
> role. </P
><DIV
CLASS="NOTE"
><BLOCKQUOTE
CLASS="NOTE"
><P
><B
>Note: </B
> The problem here requiring that you <SPAN
CLASS="QUOTE"
>"shoot the
failed node in the head"</SPAN
> is not fundamentally about replication
or <SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
>; <SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
> handles this all reasonably gracefully, as once
the node is marked as failed, the other nodes will <SPAN
CLASS="QUOTE"
>"shun"</SPAN
>
it, effectively ignoring it.  The problem is instead with
<SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>your application.</I
></SPAN
> Supposing the failed node can
come back up sufficiently that it can respond to application requests,
<SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>that</I
></SPAN
> is likely to be a problem, and one that
hasn't anything to do with <SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
>.  The trouble is if there are two
databases that can respond as if they are <SPAN
CLASS="QUOTE"
>"master"</SPAN
>
systems. </P
></BLOCKQUOTE
></DIV
><P
> When failover occurs, there therefore needs to be a mechanism
to forcibly knock the failed node off the network in order to prevent
applications from getting confused.  This could take place via having
an SNMP interface that does some combination of the following:</P
><P
></P
><UL
><LI
><P
> Turns off power on the failed server. </P
><P
> If care is not taken, the server may reappear when system
administrators power it up. </P
></LI
><LI
><P
> Modify firewall rules or other network configuration
to drop the failed server's IP address from the network. </P
><P
> If the server has multiple network interfaces, and therefore
multiple IP addresses, this approach allows the
<SPAN
CLASS="QUOTE"
>"application"</SPAN
> addresses to be dropped/deactivated, but
leave <SPAN
CLASS="QUOTE"
>"administrative"</SPAN
> addresses open so that the server
would remain accessible to system administrators.  </P
></LI
></UL
></DIV
><DIV
CLASS="SECT2"
><H2
CLASS="SECT2"
><A
NAME="REBUILDNODE1"
>3.4.7. After Failover, Reconfiguring
Former Origin</A
></H2
><A
NAME="AEN1383"
></A
><P
> What happens to the failed node will depend somewhat on the
nature of the catastrophe that lead to needing to fail over to another
node.  If the node had to be abandoned because of physical destruction
of its disk storage, there will likely not be anything of interest
left.  On the other hand, a node might be abandoned due to the failure
of a network connection, in which case the former
<SPAN
CLASS="QUOTE"
>"provider"</SPAN
> can appear be functioning perfectly well.
Nonetheless, once communications are restored, the fact of the
<TT
CLASS="COMMAND"
>FAIL OVER</TT
> makes it mandatory that the failed node
be abandoned. </P
><P
> After the above failover, the data stored on node 1 will
rapidly become increasingly out of sync with the rest of the nodes,
and must be treated as corrupt.  Therefore, the only way to get node 1
back and transfer the origin role back to it is to rebuild it from
scratch as a subscriber, let it catch up, and then follow the
switchover procedure.</P
><P
> A good reason <SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>not</I
></SPAN
> to do this automatically
is the fact that important updates (from a
<SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>business</I
></SPAN
> perspective) may have been
<TT
CLASS="COMMAND"
>commit</TT
>ted on the failing system.  You probably want
to analyze the last few transactions that made it into the failed node
to see if some of them need to be reapplied on the <SPAN
CLASS="QUOTE"
>"live"</SPAN
>
cluster.  For instance, if someone was entering bank deposits
affecting customer accounts at the time of failure, you wouldn't want
to lose that information.</P
><DIV
CLASS="WARNING"
><P
></P
><TABLE
CLASS="WARNING"
BORDER="1"
WIDTH="100%"
><TR
><TD
ALIGN="CENTER"
><B
>Warning</B
></TD
></TR
><TR
><TD
ALIGN="LEFT"
><P
> It has been observed that there can be some very
confusing results if a node is <SPAN
CLASS="QUOTE"
>"failed"</SPAN
> due to a
persistent network outage as opposed to failure of data storage.  In
such a scenario, the <SPAN
CLASS="QUOTE"
>"failed"</SPAN
> node has a database in
perfectly fine form; it is just that since it was cut off, it
<SPAN
CLASS="QUOTE"
>"screams in silence."</SPAN
> </P
><P
> If the network connection is repaired, that node could
reappear, and as far as <SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>its</I
></SPAN
> configuration is
concerned, all is well, and it should communicate with the rest of its
<SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
> cluster. </P
><P
> In <SPAN
CLASS="emphasis"
><I
CLASS="EMPHASIS"
>fact</I
></SPAN
>, the only confusion taking place
is on that node.  The other nodes in the cluster are not confused at
all; they know that this node is <SPAN
CLASS="QUOTE"
>"dead,"</SPAN
> and that they
should ignore it.  But there is not a way to know this by looking at
the <SPAN
CLASS="QUOTE"
>"failed"</SPAN
> node.</P
><P
> This points back to the design point that <SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
> is not a
network monitoring tool.  You need to have clear methods of
communicating to applications and users what database hosts are to be
used.  If those methods are lacking, adding replication to the mix
will worsen the potential for confusion, and failover will be a point
at which there is enormous potential for confusion. </P
></TD
></TR
></TABLE
></DIV
><P
> If the database is very large, it may take many hours to
recover node1 as a functioning <SPAN
CLASS="PRODUCTNAME"
>Slony-I</SPAN
> node; that is another reason
to consider failover as an undesirable <SPAN
CLASS="QUOTE"
>"final
resort."</SPAN
></P
></DIV
><DIV
CLASS="SECT2"
><H2
CLASS="SECT2"
><A
NAME="AEN1411"
>3.4.8. Planning for Failover</A
></H2
><P
> <A
HREF="failover.html"
> Failover </A
> policies
should be planned for ahead of time.  </P
><P
> Most pointedly, any node that is expected to be a failover
target must have its subscription(s) set up with the option
<TT
CLASS="COMMAND"
>FORWARD = YES</TT
>.  Otherwise, that node is not a
candidate for being promoted to origin node. </P
><P
> This may simply involve thinking about what the priority lists
should be of what should fail to what, as opposed to trying to
automate it.  But knowing what to do ahead of time cuts down on the
number of mistakes made.</P
><P
> At Afilias, a variety of internal [<SPAN
CLASS="CITATION"
>The 3AM Unhappy
DBA's Guide to...</SPAN
>] guides have been created to provide
checklists of what to do when certain <SPAN
CLASS="QUOTE"
>"unhappy"</SPAN
> events
take place.  This sort of material is highly specific to the
environment and the set of applications running there, so you would
need to generate your own such documents.  This is one of the vital
components of any disaster recovery preparations.</P
></DIV
></DIV
><DIV
CLASS="NAVFOOTER"
><HR
ALIGN="LEFT"
WIDTH="100%"><TABLE
SUMMARY="Footer navigation table"
WIDTH="100%"
BORDER="0"
CELLPADDING="0"
CELLSPACING="0"
><TR
><TD
WIDTH="33%"
ALIGN="left"
VALIGN="top"
><A
HREF="ddlchanges.html"
ACCESSKEY="P"
>Prev</A
></TD
><TD
WIDTH="34%"
ALIGN="center"
VALIGN="top"
><A
HREF="index.html"
ACCESSKEY="H"
>Home</A
></TD
><TD
WIDTH="33%"
ALIGN="right"
VALIGN="top"
><A
HREF="advanced.html"
ACCESSKEY="N"
>Next</A
></TD
></TR
><TR
><TD
WIDTH="33%"
ALIGN="left"
VALIGN="top"
>Database Schema Changes (DDL)</TD
><TD
WIDTH="34%"
ALIGN="center"
VALIGN="top"
><A
HREF="administration.html"
ACCESSKEY="U"
>Up</A
></TD
><TD
WIDTH="33%"
ALIGN="right"
VALIGN="top"
>Advanced Concepts</TD
></TR
></TABLE
></DIV
></BODY
></HTML
>
slony1-2-doc 2.2.5-2 / usr / share / doc / slony1-2-doc / adminguide / failover.html
