Re: Define a job to run once an year

joel-garry_at_home.com wrote:

>
> I have seen very few systems stable enough that I would consider
> defining a job to run once a year. (The OP was really asking how to
> run a job at the end of the year). I have seen screwups on systems
> where a job was only supposed to run once and then ran again the next
> year :-O (cron without full time sys admins seems particularly
> vulnerable to that).
>
> jg
> --
> @home.com is bogus.
> More than 2,500 left-handed people are killed annually from using
> products made for right-handed people.
>

I had the same misgiving -- system time is a very vulnerable to errors,

   even with NTP (network time protocol) it seems many sites don't implement it with the minimum required level of robustness (3 independent stratum 1 sources). I've seen entire corporate networks slip by as much as 6 hours over a quiet weekend due to time synchronization problems, very hard to monitor within band.

Despite the discussion of real-life requirements for sequence numbers, the problem here seems to be that all transactions (started? completed?) prior to midnight must use one sequence, and those after must use another. If this is truly critical for data integrity, then I would argue an outage ("quiesce database", for example) should be taken to drop and recreate the sequence, not just a background scheduled job.   Probably calls for an on-call DBA to intervene by hand.

There is also a link for AskTom

http://asktom.oracle.com/pls/ask/f?p=4950:8:5358747439817317758::NO::F4950_P8_DISPLAYID,F4950_P8_CRITERIA:1119633817597

which shows how to reset a sequence without dropping/recreating it.

-Mark Bole Received on Wed Dec 29 2004 - 14:41:32 CST