Re: HELP: Long (lasting) transactions?

In article <3n06td$3pi@sword.eng.pyramid.com>, graeme@pyramid.com (Graeme Sargent) writes: |>Roy F. Swonger (swonger@kwalty.enet.dec.com) wrote: [a few lines of previous conversation deleted] |> |>: In general, I'd think that such long transactions would present a problem for |>: recoverability in any database system. It's sort of like backing up your |>: system disks; if you only do it every few days, can you really afford to lose |>: a few days' worth of data? With transactions, if some catastrophe occurs three |>: days into a transaction, can you really afford to have those three days of |>: activity be completely lost? |> |>Just because a transaction is long-lasting does not mean it is |>intensive. If the three days of activity consisted of three singleton |>updates, then it could be perfectly acceptable to have to redo it. |>There is no reason why any other activity during those three days should |>need to be affected. Yes, it *could* be acceptable to re-do the lost work. My point was that the original noter should consider the question. After all, things other than the level of activity may be important; perhaps the transaction uses data that would not be recoverable, like logging current timestamps in a history relation, or taking interactive user input. Or, it may be a long read-only transaction, in which case there's no problem. Or, anything in between. Like anything else, the user must consider the specific situation, its risks, and potential benefits. Roy -- -------- Roy F. Swonger Oracle Rdb Software Quality Engineering swonger@nova.enet.dec.com The opinions expressed in my notes are solely my own and not those of my employer. --------