[Bucardo-general] BYTEA columns not synchronized correctly

Michelle Sullivan michelle at sorbs.net
Thu Mar 1 20:57:36 UTC 2012


Christian Kaltepoth wrote:
> Dear Bucardo team,
> 
> first of all thank you for all your work on this great software. I just
> started to use Bucardo but I'm really impressed how great everything
> works.
> 
> But unfortunately we have one major problem. We have on table containing
> a BYTEA column that doesn't sync correctly using 'swap' mode. Let me try
> to describe the problem.
> 
> If we insert a new row into the corresponding table the data looks like
> this:
> 
> \x30820247308201b0020900a82498c179eef339300d06092a864886f70d0101050
> 5003068310b3009060355040613024445310c300a060355040813034e5257311130
> 0f06035504071308446f72746d756e643110300e060355040a1307696e67656e697
> 431143012060355040b130.....
> 
> This data is correct. I manually decoded the HEX representation and
> confirmed that the data is exactly the data we wrote to the database.
> 
> The column is then automatically duplicated to the second host, but on
> the second host the data gets messed up. The row on the second host
> contains completely different binary data:
> 
> \x78333038323032343733303832303162303032303930306138323439386331373
> 9656566333339333030643036303932613836343838366637306430313031303530
> 3530303330363833313062333030393036303335353034303631333032343434353
> 33130633330306130363033353530...
> 
> We used the psql command line client on both hosts to get this HEX
> representation out of the database.
> 
> Two things about the broken data is eye catching:
> 
>       * The total length of the broken HEX representation on the second
>         host is twice as long as the correct HEX representation on the
>         first host.
>       * The broken data doesn't contain a single letter character (A-F)
>         which is very strange for a HEX representation.
> 
> Do you have any idea what is causing this?
> 
> Some details about my setup.
> 
>       * CentOS 6.2 Minimal Install
>       * Bucardo 4.4.8
>       * PostgreSQL 9.0.7
> 
> Thanks for your help


I'm just going to throw something out there ...  have you checked the
default character sets on all systems involved?  I have seen issues with
various (not Bucardo specifically, eg I've seen it with MySQL as well)
systems where not having the default charset has resulted in differing
data sets that is supposed to be a 'replica' ... sometimes it is the
data different, sometimes it is the console displaying it and the data
is actually correct.

Michelle


-- 
Vulnerabilities are weaknesses associated with an organisations assets
that maybe exploited by a threat causing unwanted incidents.
http://www.mhix.org/


More information about the Bucardo-general mailing list