Archiving And Temporary Spaces

Problem: Temp Space Problems

I'm archiving very large 24 x 7 databases while they are seeing significant amounts of user updates, inserts, and deletes. My archive is failing with temp space errors. Why?

Solution: Use a Big Temporary Space

This is similar to the quandary of one large DBSPACETEMP versus a DBSPACETEMP composed of more, smaller dbspaces. While you want to have multiple smaller DBSPACES comprising your DBSPACETEMP for query performance, IDS versions before 7.3 have some problems with DBSPACETEMP usage.

IDS systems changed their archiving methodology from OnLine. Now, at each checkpoint during an archive, the physical log pages are stored in one of the tempdbspaces listed in DBSPACETEMP before the physical log is cleared and the checkpoint is allowed to complete. When the archive of each dbspace is completed, these temp tables are written out to tape as part of the archive. This is the mechanism that allows IDS to simultaneously archive to tape and to modify pages that may have already been written to tape.

These temp tables are distributed across all of the dbspaces in DBSPACETEMP, but each temp table will not span dbspaces. Thus, under very heavy user activity, a temp dbspace holding the stored physical log spaces can fill up, even though there is plenty of space left in the other temp dbspaces. When this happens, the archive aborts. This is made worse by the fact that versions prior to 7.3 do not clean out these dbspaces after each regular dbspace is archived. This occurs only after the entire archive is completed.

The way to cure this is to configure your DBSPACETEMP as one or two multichunk, large dbspaces rather than as many small dbspaces. This is scheduled to be fixed in 7.3.