Dear Friends:

I have always been under the impression that if you had a table with a column that was TIMESTAMP NOT NULL WITH DEFAULT - and you never provided a value for that column but rather took the default for all inserts, that you would be guaranteed uniqueness. I was told that such a column would be a perfect Primary Key if there were no "natural" primary key - because it was guaranteed unique. I am fairly sure I remember an instructor saying, only partly as a joke, that DB2 could *NEVER* insert two rows in the same millionth of a second.

Now I am looking at a situation involving a 280,000,000 row production table with such a column. There are 3 situations on one day - within a 5 minute period - where I have 2 duplicates down to the microsecond, though the rest of the data is totally different - just the timestamps are the same.

Does anyone have any theories how this could happen? I realize, of course, I could create a unique index to enforce uniqueness, but I am surprised this is necessary.

Best, Hal

PS: No replication is involved, no Data Partitioning either. This is a single instance - DB2 V8.2.3


DB2 & UDB email list listserv db2-l LazyDBA home page