So I have a massive ssis load that I have to do daily. It is all droptable/createtable/load/index type stuff. It loads 69 tables, named XXX_loading, along side the normal table names, then when everything is complete, it drops the normal tables, and renames all the indexes and tables, to replace the normal table names, and prep for the next night of loads. total db size when done is 834gb, with 495gb free, which is expected, since it doubles the database during the load, and then drops the original tables.
When executed as a sql agent job, It randomly hits certain table loads, and gives the wait type PREEMPTIVE_SHAREDMEM_GETDATA, and grinds the package to a halt, during that table's data flow. I've tried breaking them up into smaller jobs, and running them in sequence, which helped it get a little further, but it is still hitting somewhere in the packages, and stops the succession.
I'm on Microsoft SQL Server 2008 R2 (RTM) - 10.50.1777.0 (X64) - which is sp1 with CU7. I see that there is a sp2, but it has only been out for a few days, and I'm hesitant.
During testing, running from the IDE in BIDS, this issue never cropped up. It only shows up when I deploy the package, and execute it as a sql agent job. Upon this coming up, I can drop right to BIDS, and run the package manually without issue, but it sucks to have to do that at 3am. Man needs his beauty sleep!
I've eliminated all the running items on the server that occur during the time span of this job, like backups, other data loads, etc.
I thought I had it licked when I broke them up into smaller ssis packages, and it ran fine for two days, but now its back. The only other thing I did differently for those two days is restart sql agent a little while before the package string started. Sure, I can set sql agent to be restarted somehow, but I'd rather fix the issue at its root.
Anybody else ever get this behavior? is there a workaround? quick fix? CU#?