On 9/12/2007 3:40 PM, JP Fletcher wrote:
> I just noticed when doing a different cluster, that the same thing seems 
> to happen with STORE NODE

It is a side effect of that TABLE ADD KEY misfeature. It requires a 
cluster wide uniquely maintained sequence, and the way slony does that 
is by setting that sequence to the nodeID * 1000000000000000.

I didn't plan for someone using birth dates or lottery ticket numbers as 
nodeID's ;-) Sorry, for the time being you have to stay below something 
around 9220 ... the problem will be gone in 2.0.


Jan

> 
> /opt/scripts/slonik/cluster/cluster_init.slonik:5: PGRES_FATAL_ERROR 
> select "_cluster".initializeLocalNode(650901, 'Node 650901'); select 
> "_cluster".enableNode_int(650901);  - ERROR:  bigint out of range
> CONTEXT:  SQL statement "SELECT  setval('"_cluster".sl_rowid_seq',  $1 
> ::int8 * '1000000000000000'::int8)"
> PL/pgSQL function "initializelocalnode" line 26 at perform
> 
> The slonik command was
> 
> STORE NODE (ID = 650901, EVENT NODE = 22, COMMENT = 'Node 650901');
> 
> These node ids may have been created initially on 1.0.5, many months ago...
> 
> JP Fletcher wrote:
>> Hi,
>>
>> I'm trying to create a  test cluster in the image of one that already 
>> exists.  The script I'm using  connects to the existing cluster, then 
>> builds an INIT CLUSTER command based on the current origin node.  The 
>> id of the node in question is 40003.  The INIT CLUSTER fails with the 
>> following:
>>
>> /opt/scripts/slonik/cluster/cluster_init.slonik:2: PGRES_FATAL_ERROR 
>> select "_cluster".initializeLocalNode(40003, 'Node 40003'); select 
>> "_cluster".enableNode(40003);  - ERROR:  bigint out of range
>> CONTEXT:  SQL statement "SELECT  setval('"_cluster".sl_rowid_seq',  $1 
>> ::int8 * '1000000000000000'::int8)"
>> PL/pgSQL function "initializelocalnode" line 26 at perform
>>
>>
>> The node in question was added via STORE NODE, and a MOVE SET was done 
>> at some time.  If a node id is acceptable to STORE NODE should it not 
>> also be acceptable to INIT CLUSTER?  I happened to notice this on a 
>> 1.1.5 cluster, but it looks to me that initializeLocalNode() is 
>> identical in v 1.2.
>>
>> JP
>>
> 
> 


-- 
#======================================================================#
# It's easier to get forgiveness for being wrong than for being right. #
# Let's break this rule - forgive me.                                  #
#================================================== JanWieck at Yahoo.com #