Re: [Treebase-devel] Hibernate_sequence need a bigger value

[Vladimir G. <vla...@du...>]

On Dec 25, 2009, at 11:11 AM, youjun guo wrote:

> method 1. use one single sequence for all the domain table;
> method 2. assigned each domain table a induvadule sequnce.

I would be in favor of using per-table postgresql sequences to  
generate IDs.  I just think that, whatever solution we are inclined  
for, we should first make a good effort to understand what is going  
on.  E.g.,

- As Bill alludes, different tables/objects might have been set up to  
use different methods of sequence generation.  Then, our fix might  
have to be different for each method.
-  Is it possible that the single hibernate_sequence for all stored  
objects was used to ensure that there is a unique identifier for each  
stored object?  Is there code that may depend on this? (e.g., some  
kind of audit?)


> Also, for method 2, from now on, all these 70 sequence need to  
> maintain the synchronizing with data insert and data migration of  
> their table. I found some of them already lose the synchronise .

I would guess the loss of id synch has happened for those tables xxx  
whose xxx_id_sequence value  after TB1 import was below the  
hibernate_sequence value.  The remaining tables (as you encountered  
for taxonlabel and phylotreenode) just refused to insert new rows  
(since ids generated for them by hibernate_sequence were clashing with  
existing rows).

--Vladimir