Replicating LOBs

[c***@us.ibm.com]

Hi.

I'm trying to replicate some tables that contain LOBs (three have BLOBs and one has a CLOB) between two DB2 servers. Both servers are running AIX 5.3, but one has DB2 v8.1 FP12 and the other has DB2 v8.1 FP11. One of the LOB-holding tables replicates just fine (a BLOB), but the other three fail with the following errors being posted in the replication load log.

SQL3168W The PC/IXF column specified to be loaded into database column "3" is
not compatible with the database column. Nulls will be inserted.

SQL3088N The source column specified to be loaded into database column "4" is
not compatible with the database column, but the database column is not
nullable.

SQL3088N The source column specified to be loaded into database column "4" is
not compatible with the database column, but the database column is not
nullable.

I have specified that the LOB(s) must be replicated (in DB2RC) and mapped it to the appropriate target attribute(s). I am puzzled that one works, where the other three fail.

The table which succeeds has the LOB as the last attribute in the table. Could it be that the table designs need to have the LOB as the last attribute? I would not think so, but the errors that I'm getting are making me think that the full refresh is not doing the EXPORT/LOAD correctly.

[c***@us.ibm.com]

Well, I ran a test replicating a version of one of the tables where the LOB was relocated so it was the last attribute on the table. I created the table, loaded it with data, registered it, and subscribed to it with a test set. After modifying the set member so that the LOB was included, and mapping it to a target table (that had a matching layout), I was able to run a full refresh without difficulty. No errors came in stating there were errors with coordinating columns (as before).

This is the first time I attempted to replicate a table which contained a LOB elsewhere in the attribute list, and it seems pretty peculiar. The other LOB containing table that I replicate has the LOB as the last attribute by design, and had never posed a problem before. There seems to be a problem with the position of the LOB within the table.

So, I have revised the three tables so that they have their LOB as the last attribute and reworked my subscription set. No longer a problem.

Anyone from Toronto Lab or Germany Lab have an explanation as to why the export/load function was getting confused when the LOB was anywhere in the table except the last attribute?

Thanks.