>> I have 2 nodes, node 1 master, node 2 slave. Replication works  fine
>> (with postgresql 8.0.3, slony 1.1.0)
>> Suppose node 1  is down.
>> I want to promote node 2 to new master, so I follow the  procedure :
>> step 1 : failover, ok
>> step 2 : drop node 1,  ok
>> step 3 : store node 1 (new slave), store path and store listen,  ok
>> step 4 : subscribe node 1 as receiver for the set, provider is  node 2,
>> ERROR : fe_sendauth : no password  supplied
>>
>> has anyone the same problem  ?
>>
>> -----
>> my node configuration  :
>>
>> first master :
>> node 1 localhost  5432
>> node 2 localhost 5500
>>
>> first slave  :
>> node 1 localhost 5500
>> node 2 localhost  5432
>>
>> the 2 databases communicate via ssh tunneling (port  5500 to 5432).

>Is it the slonik script that fails?  Or is it  slon that fails?  That's
>not clear here...
 
It's the slonik script. 

>In any case, the "Best Practices"  handling of passwords is to store them
>in $HOME/.pgpass
 
That's what I do.

>Slony-I doesn't do anything at all special  with the handling of
>authentication; it uses libpq, and hence if you are  having connection
>problems, you should try using the same configuration  to connect using
>psql.  If psql asks for a password, that would be a  hint that you need
>to stuff a password into $HOME/.pgpass
 
I tried the same failover/subscribe process after exchanging the 2 servers,  
and I have the same problem.
 
but actually it think there is a config problem on one of the 2  servers as I 
found that the standard php pg_connect function fails with  host=localhost 
and succeed with host='' (empty). very strange.
 
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://gborg.postgresql.org/pipermail/slony1-general/attachments/20050810/eec61e25/attachment.html