On 2/1/2007 9:39 AM, Andrzej Kwiatkowski wrote:
> This is my first post here, so greets for all.
> And my problem with failover.
>
> Idea of my tests is:
> 1,2 master & slave for database (switched in case of failuer)
> 3 - replica for some reports and so on.
>
> I'm starting my test slony cluster with:
>
>
>
> cluster name = test;
>
> node 1 admin conninfo = 'dbname=isp host=localhost port=5432 user=pgsql';
> node 2 admin conninfo = 'dbname=isp host=localhost port=6000 user=pgsql';
> node 3 admin conninfo = 'dbname=isp host=localhost port=6001 user=pgsql';
>
> init cluster ( id = 1, comment= 'MASTER');
>
> store node (id=2, comment = 'SLAVE');
> store node (id=3, comment = 'REPLICA');
>
> # between master
> store path (server = 1, client = 2, conninfo ='dbname=isp host=localhost
> port=5432 user=pgsql' );
> store path (server = 2, client = 1, conninfo ='dbname=isp host=localhost
> port=6000 user=pgsql' );
>
> # from master to replica
> store path (server = 1, client = 3, conninfo ='dbname=isp host=localhost
> port=5432 user=pgsql' );
> store path (server = 2, client = 3, conninfo ='dbname=isp host=localhost
> port=6000 user=pgsql' );
>
> # from replica to master
> store path (server = 3, client = 1, conninfo ='dbname=isp host=localhost
> port=6001 user=pgsql' );
> store path (server = 3, client = 2, conninfo ='dbname=isp host=localhost
> port=6001 user=pgsql' );
>
> #
> # between masters
> store listen (origin=1, receiver=2, provider=1);
> store listen (origin=2, receiver=1, provider=2);
>
> # from master to replica
> store listen (origin=1, receiver=3, provider=1);
> store listen (origin=2, receiver=3, provider=1);
> store listen (origin=1, receiver=3, provider=2);
> store listen (origin=2, receiver=3, provider=2);
>
> # from replica to masters
> store listen (origin=3, receiver=1, provider=3);
> store listen (origin=3, receiver=2, provider=1);
>
> store listen (origin=3, receiver=1, provider=2);
> store listen (origin=3, receiver=2, provider=3);
>
> create set (id=1, origin=1, comment='FOR SLAVE');
> set add table (set id = 1, origin = 1, id = 1, full qualified name =
> 'public.test', comment = '');
> set add sequence (set id = 1, origin = 1, id = 2, full qualified name =
> 'public.test_id_seq', comment = '');
>
> subscribe set (id=1, provider=1, receiver=2, forward=yes);
> subscribe set (id=1, provider=1, receiver=3, forward=no);
>
>
> and then starting slon daemons.
>
> Everything is ok, replication is done, moving master is ok
> with lock and move set.
>
> But when i kill postmaster and slon daemon for database id 1
> and execute script
>
> cluster name = test;
>
> node 1 admin conninfo = 'dbname=isp host=localhost port=5432 user=pgsql';
> node 2 admin conninfo = 'dbname=isp host=localhost port=6000 user=pgsql';
> node 3 admin conninfo = 'dbname=isp host=localhost port=6001 user=pgsql';
>
> ### MOVE FROM SLAVE TO MASTER
> #
>
> failover (id=1, backup node = 2);
> drop node (id = 1, event node = 2);
Do not drop the node instantly after issuing the failover. Wait until
all other nodes have finished processing the failover and accept set.
Jan
>
> with output:
>
> <stdin>:10: NOTICE: failedNode: set 1 has other direct receivers -
> change providers only
> <stdin>:10: NOTICE: failedNode: set 1 has other direct receivers -
> change providers only
> IMPORTANT: Last known SYNC for set 1 = 21
>
>
> Nothing is done, and slon daemon output looks like:
>
> 2007-02-01 15:36:11 CET DEBUG2 syncThread: new sl_action_seq 1 - SYNC 14
> 2007-02-01 15:36:11 CET DEBUG2 remoteListenThread_2: queue event 2,15
> ACCEPT_SET
> 2007-02-01 15:36:11 CET DEBUG2 remoteListenThread_2: queue event 2,16
> DROP_NODE
> 2007-02-01 15:36:11 CET DEBUG2 remoteWorkerThread_2: Received event 2,15
> ACCEPT_SET
> 2007-02-01 15:36:11 CET DEBUG2 start processing ACCEPT_SET
> 2007-02-01 15:36:11 CET DEBUG2 ACCEPT: set=1
> 2007-02-01 15:36:11 CET DEBUG2 ACCEPT: old origin=1
> 2007-02-01 15:36:11 CET DEBUG2 ACCEPT: new origin=2
> 2007-02-01 15:36:11 CET DEBUG2 ACCEPT: move set seq=22
> 2007-02-01 15:36:11 CET DEBUG2 got parms ACCEPT_SET
> 2007-02-01 15:36:11 CET DEBUG2 ACCEPT_SET - node not origin
> 2007-02-01 15:36:11 CET DEBUG2 ACCEPT_SET - MOVE_SET or FAILOVER_SET not
> received yet - sleep
> 2007-02-01 15:36:13 CET DEBUG2 remoteListenThread_2: queue event 2,17 SYNC
> 2007-02-01 15:36:17 CET DEBUG2 localListenThread: Received event 3,14 SYNC
> 2007-02-01 15:36:19 CET ERROR slon_connectdb: PQconnectdb("dbname=isp
> host=localhost port=5432 user=pgsql") f
> ailed - could not connect to server: Connection refused
> Is the server running on host "localhost" and accepting
> TCP/IP connections on port 5432?
> 2007-02-01 15:36:19 CET WARN remoteListenThread_1: DB connection
> failed - sleep 10 seconds
> 2007-02-01 15:36:19 CET DEBUG2 remoteListenThread_2: LISTEN
>
>
>
>
>
>
>
> id databases in namespace _test
>
> on replica:
> SELECT * from _test.sl_node
> isp-# ;
> no_id | no_active | no_comment | no_spool
> -------+-----------+------------+----------
> 1 | t | MASTER | f
> 2 | t | SLAVE | f
> 3 | t | REPLICA | f
>
>
> on slave:
>
> isp=# SELECT * from _test.sl_node
> isp-# ;
> no_id | no_active | no_comment | no_spool
> -------+-----------+------------+----------
> 2 | t | SLAVE | f
> 3 | t | REPLICA | f
> (2 rows)
>
>
> It looks like failover comment isn't properly propagated to replica.
> I was tested a lot of scenarios, but all failed...
>
> Thanks for help.
> AK
> _______________________________________________
> Slony1-general mailing list
> [email protected]
> http://gborg.postgresql.org/mailman/listinfo/slony1-general
--
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me. #
#================================================== [EMAIL PROTECTED] #
_______________________________________________
Slony1-general mailing list
[email protected]
http://gborg.postgresql.org/mailman/listinfo/slony1-general
