Re: DB2 Auto-update TimeStamp

On 11-Jun-2014 09:32 -0500, Charles Wilt wrote:

I seem to recall hearing that the DB was smart enough to not do an
update if nothing actually changed.

I seem to recall that the journal images would not be produced for an update of no change. But I do not recall that the database would ever dismiss such an update request; the OS DB certainly does not dismiss a request, but the LIC DB could [just as I believe it does for when the image is not passed to effect an R-UP being deposited in the journal [receiver].

Though there appears to be some caveats to that. I tested a table
with a row change timestamp and the TS was changed.

If the LIC DB does ignore an update method if the data does not change, then that may possibly be available only with commitment-control\isolation being other than NC; i.e. perhaps an alternate test to try.

I'd assume that a before/after update trigger would also force the
DB to make the change. <<SNIP>>

A trigger is really moot in that regard, except one that signals an error.? That is, the UPDATE was still requested, and any trigger that does not prevent the I\O will let that update request complete just as if there were no trigger(s); whether a before-event trigger had updated the row image to be different or the same as the before image would seem to be the deciding factor with regard to whether the /change/ occurs at the LIC level.