On 4/21/2010 7:06 PM, Jaime Casanova wrote:
> On Wed, Apr 21, 2010 at 8:22 AM, Steve Singer <ssinger at ca.afilias.info> wrote:
>>
>> http://bugs.slony.info/bugzilla/show_bug.cgi?id=111 mentions this 'node -1'
>> error, I tried to reproduce this with the 2.0 branch a while back but wasn't
>> able to figure out exactly what steps I need to perform to reproduce this
>> state.
>>
>> If your in a position that you can make a pg_dump of your slony schema (from
>> both the master and the replica, with the data in the sl_* tables) and
>> somehow send it to us off-list we wouldn't mind taking a look at it.
>>
> 
> actually is too late for this, the schema has more than 800Mb now...
> for some reason sl_log_2 has grown up to 3147101 rows, so i executed
> logswitch_start() and logswitch_finish() and get this message:
> """
> NOTICE:  Slony-I: log switch to sl_log_1 still in progress - sl_log_2
> not truncated
>  logswitch_finish
> ------------------
>                -1
> (1 row)
> """
> 
> and now sl_log_2 is not growing any more, but now it's sl_log_1 who is
> grown... but i can't find any messages on postgres or slony logs
> explaining why

Once replication for a node has stalled and cannot get past a certain 
event, no log switch will ever succeed. Since you keep getting that node 
-1 error message, I presume you are in exactly that situation. 
logswitch_finish() cannot succeed because that node will never consume 
all the log rows, that are in that segment. This is not a bug, this is 
actually a well designed feature.

> maybe now that 2.0.3 is released i will uninstall 1.2.20 and install
> the new one... but i prefer to find an answer to this...
> i will try to make the same bug trigger on my home machine, hope i can :)
> 
> any ideas on what look for the sl_log_x problem?

What does sl_status look like on the origin?


Jan

-- 
Anyone who trades liberty for security deserves neither
liberty nor security. -- Benjamin Franklin