Обсуждение: Who is Slony Master/Slave + general questions.

Hello,

  I'm starting to use slony as a redundancy solution for the project
I'm currently working on.  Running SuSE Linux 9 where one machine
contains the prime database and the second machine contains the backup
database.  The Slony version I'm using is 1.1.2.  If some of the issues
have been addressed in the newer version of Slony, please let me know.

I have looked at the Nagios scripts and others and am still left with
questions regarding how to dynamically determine who is slave and who
is master during normal and failover operations.  Take a scenario that
you want to check the state of the system without prior knowledge of
the node setup, how would you determine which machine is the prime and
which one is the slave?

Also I'm having issues with the slonik script (below) that is supposed
to handle the failover to the slave in case of master failure.  For
some reason it hangs and I was wondering if there are known issues with
it.  The test condition I'm working with is: reboot the master, the
slave is supposed to take over.

slonik <<_EOF_
    # ----
    # This defines which namespace the replication system uses
    # ----
    cluster name = $CLUSTER;

    # ----
    # Admin conninfo's are used by the slonik program to connect
    # to the node databases.  So these are the PQconnectdb arguments
    # that connect from the administrators workstation (where
    # slonik is executed).
    # ----
    node 1 admin conninfo = 'dbname=$DBNAME1 host=$HOST1 port=5432
user=$SLONY_USER1';
    node 2 admin conninfo = 'dbname=$DBNAME2 host=$HOST2
user=$SLONY_USER2';

    # ----
    # Node 2 subscribes set 1
    # ----
    failover ( id = 1, backup node = 2);
_EOF_


Thanks a lot for your help,

Slawek

Hello,

You should ask directly to the slony1 mailing list.

sjarosz@gmail.com a écrit :
> (...) The Slony version I'm using is 1.1.2.
The current version of Slony1 is slony1-1.2.6.
> Take a scenario that
> you want to check the state of the system without prior knowledge of
> the node setup, how would you determine which machine is the prime and
> which one is the slave?
>
Without any knowledge of replication ? That will be difficult. You
should connect to one of the DB and have a look at slony schema tables
(sl_status and sl_listen for instance)...

You may also have a look at that page :
http://linuxfinances.info/info/monitoring.html

> Also I'm having issues with the slonik script (below) that is supposed
> to handle the failover to the slave in case of master failure.  For
> some reason it hangs and I was wondering if there are known issues with
> it.
As written in documentation "Slony-I does not provide any automatic
detection for failed systems. "
First of all, you may want to upgrade to the latest stable slony1 version.


Cheers,

SAS

>>As written in documentation "Slony-I does not provide any automatic
>>detection for failed systems. "
>>First of all, you may want to upgrade to the latest stable slony1 version.

But if you have a combination of Slony + Linux HA you can make use of Slony failover to do an automatic failover when the master node goes down.

--------------
Shoaib Mir
EnterpriseDB ( www.enterprisedb.com)

I am using LinuxHA to manage the failover and Slony as part of to
failover to move to the healthy node.  But my question was more along
the lines, if a user has access to both databases (master and slave)
but does not know which one is which, how can you tell?

Take a scenario: you configure 2 servers as master and slave.  You walk
for a period of time during which a number failovers occur.  You come
back.  Can I query a sl_???? table to determine which server is the
current master and which one is the current slave?

Thank you,

Slawek

sjarosz@gmail.com writes:
> Take a scenario: you configure 2 servers as master and slave.  You walk
> for a period of time during which a number failovers occur.  You come
> back.  Can I query a sl_???? table to determine which server is the
> current master and which one is the current slave?

In a sense, the question is a bad one.

There is nothing about a server which inherently gives it a role as
either master or slave as far as Slony-I is concerned.  Nodes are just
nodes.

You may determine that a particular node is the origin of some
particular replication set; that would indicate that, with respect to
that set of tables, a particular node is "master."

Look at sl_set; it contains a list of sets, and indicates, for each,
which node is the origin.
--
"cbbrowne","@","acm.org"
http://cbbrowne.com/info/languages.html
Rules of the Evil Overlord #133.  "If I find my beautiful consort with
access to  my fortress has been  associating with the  hero, I'll have
her executed.  It's regrettable,  but new consorts  are easier  to get
than new fortresses  and maybe the next one will  pay attention at the
orientation meeting." <http://www.eviloverlord.com/>

I dont have the replication setup on my machine right now but I guess as far as I remember you can surely check for the master and slave nodes from a Slony schema table.

------------
Shoaib Mir
EnterpriseDB ( www.enterprisedb.com)

On Sat, 20 Jan 2007 11:07:57 +0500, "Shoaib Mir" <shoaibmir@gmail.com> wrote:
> I dont have the replication setup on my machine right now but I guess as
> far
> as I remember you can surely check for the master and slave nodes from a
> Slony schema table.
>

I think the notion of "master and slave server" is a little bit misleading here:
We have sets and a node could be a origin or subscriber of them. Thinking that way,
one idea to get that information is to issue

SELECT
  a.set_id,
  a.set_comment,
  (SELECT last_value FROM _replication.sl_local_node_id) AS local_id,
  CASE WHEN a.set_origin = (SELECT last_value FROM _replication.sl_local_node_id) THEN
         TRUE
       ELSE
         FALSE END AS master_node
FROM
  _replication.sl_set a;

This gives you a result set which holds TRUE for every set the current node is
an origin node for.


> ------------
> Shoaib Mir
> EnterpriseDB (www.enterprisedb.com)
>
> On 19 Jan 2007 08:25:23 -0800, sjarosz@gmail.com <sjarosz@gmail.com>
> wrote:
>>
>> I am using LinuxHA to manage the failover and Slony as part of to
>> failover to move to the healthy node.  But my question was more along
>> the lines, if a user has access to both databases (master and slave)
>> but does not know which one is which, how can you tell?
>>
>> Take a scenario: you configure 2 servers as master and slave.  You walk
>> for a period of time during which a number failovers occur.  You come
>> back.  Can I query a sl_???? table to determine which server is the
>> current master and which one is the current slave?
>>

If you are using LinuxHA you have a virtual IP adress for your Cluster which points to the current
active "master" on your cluster. Connecting to the master node should always happen through this
IP adress, so you always "know" you are on the master when using this IP. You could then spread read
operations along the IPs assigned directly to each node, "declaring" these connections read only.

Bernd