Re: How to lock a row over a SELECT followed by an UPDATE

"Andrew Baker" <webmaster_at_vbusers.com> wrote in message news:c19b84e5.0407290256.1e869438_at_posting.google.com...
| After thinking through swapping to using a sequence number I came up
| with a couple of potential gotchas:
|
| 1. When we fail over to our DR (disaster recovery) boxes the numbers
| must continue their sequence. Would a sequence number work under this
| circumstance?

yes, the sequence definition should remained synchronized (see #2)

|
| 2. When the box reboots/crashes does it continue the sequence number
| from where it was before the restart?

lookup the information on SEQUENCE in the oracle docs, it explains how each sequence has a cache that is loaded into memory as needed, and discarded at shutdown if not used

|
| 3. Can you manually tune the sequence numbers to increment by
| specified amounts?
|

yes -- lookup CREATE SEQUENCE in the Oracle SQL manual

| 4. Is a sequence number system wide? ie. I can it be table specific
| like IDENTITY columns in SQL server?

just like tables, it depends on privileges that you grant -- lookup information about object
security in the manuals, including GRANT in the Oracle SQL manual

5) http://tahiti.oracle.com for manuals

6) your earlier post says 'it flies', referring to performance of table-based SA-ID management. maybe with one user, but if you simulate a load you will begin to see degradation due to serialization

++ mcs Received on Thu Jul 29 2004 - 14:00:31 CEST