Archives of the TeradataForum

Message Posted: Tue, 18 Nov 2003 @ 12:51:56 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Anomy,

Funny you should mention this problem...I believe my posting last week (regarding the spoiling of cache) had to do with my efforts to recreate this problem on demand.

Currently, with the recreation of JIs (and possibly the updating of JIs, I haven't verified this), a row hash write lock occurs on the DBC.TVM table for all TVMIDs (tables) assoicated with the JI build. If the TVM info resides in the user's PE cache, then a user can access these tables without any issues; however, if the TVM info does not reside in the user's PE cache, then the optimizer must get the info from DBC.TVM; however, if a JI is being built, then the respective TVMIDs are write locked. I'm pretty certain that optimizer's requests for TVM info would require a read lock (a dirty TVM read could be bad); therefore, the user will have to wait until the write lock is lifted (ie, when the JI build is complete.)

I belive Teradata is aware of this issue; however, I'm not certain when a solution is planned for implementation. As with all DRs, the more feedback NCR gets from the you (the customer), the greater the likelihood of an expedited solution. I'd recommend you contact your local NCR representative.

Thanks,

Clay

	<Prev	Next>		<<First	<Prev	Next>	Last>>