OK, had to add another layer in the form of a iTVF and then use CROSS APPLY. See if this setup works for you:

IF EXISTS ( SELECT *

...

What is the expected result from this new set of test data please?

Are you getting an error by chance or just wrong results? If you post some sample data that demonstrates why it is not working I'll try to take another look...

zach_john (5/26/2011)

---

Ninja's_RGR'us (5/25/2011)

---

I'm no SSIS guru but I know this for sure, you can commit the whole process in batches and unless you specifically ask for an ordered insert you...

try/catch is not a form of a loop...it's an exception handling structure. I prefer it over using traps because...

Hmm...running out of ideas...

Are the 3-byte "table prefix char(15)" always used in a sequential order? i.e. Could you possibly load a batch with a prefix of "xyz" before a batch...

Ninja's_RGR'us (5/25/2011)

---

opc.three (5/25/2011)

---

Ninja's_RGR'us (5/25/2011)

---

Also the LEADING part of the key is a prefix, NOT a date, NOT a sequential GUID. So that alone can cause that much fragmentation.

The only...

Ninja's_RGR'us (5/25/2011)

---

Also the LEADING part of the key is a prefix, NOT a date, NOT a sequential GUID. So that alone can cause that much fragmentation.

The only real "identity"...

OK...so much for that...

What about Ninja's thought...are the rows being updated in the table after they are initially inserted?

Lowell (5/25/2011)

---

opc.three (5/25/2011)

---

Just eyeballing your code so forgive me if this is a rabbit hole but I think if client_net_address is NULL then you end up with something that will...

Ninja's_RGR'us (5/25/2011)

---

opc.three (5/25/2011)

---

The idea of having an app generate a sequential GUID outside the database and provide it to SQL Server as the clustering key for a table is great...

<mini_rant>The use of GUID for this entity is a misnomer in the context of Microsoft SQL Server so I am calling them IDs from now on because the terminology is...

Just eyeballing your code so forgive me if this is a rabbit hole but I think if client_net_address is NULL then you end up with something that will anger most...

The idea of having an app generate a sequential GUID outside the database and provide it to SQL Server as the clustering key for a table is great on paper...but...

Agreed, and it will be exacerbated due to the 100% fill factor.