Re: Logical Standby
- From: "Mark Strickland" <strickland.mark@xxxxxxxxx>
- To: Smith.Steven@xxxxxxx
- Date: Wed, 9 May 2007 13:43:25 -0700
I'm using Logical Standby with 10.1.0.5 (RAC, Solaris 5.9, Veritas, Hitachi
SAN). I've hardly pushed the standby so I can't speak to performance
implications. For me, it has been quite stable and there's rarely a SQL
Apply crash. However, you have to be careful about large transactions in
the primary, anything that will insert/update/delete more than, say, 100,000
rows. The redo stream contains row-level DML statements that correspond to
the single DML executed in the primary. Over 100,000 such statements
and SQL Apply will eventually slow to a crawl. A DML statement that affects
1,000,000 rows would likely finish sometime during the Obama
Administration. His second term. So, if there are large transactions that
have to be performed in production (add new column to large table, set
column to default value for all rows, add not null constraint), we wrap them
in a PL/SQL block using a cursor for loop with a commit every 10,000 rows.
Works fine. The Logical Standby keeps right up. However, I can't 100% rely
on folks to do large transactions correctly, so I can't 100% rely on the
Logical Standby for any mission-critical reporting, ie, anything that
couldn't wait 24 hours while I refresh a table or the entire database. I've
had an SR open for 9 months and a bug has been filed. I did test this in
10.2.0.2 and the behavior is the same.
Regards,
Mark Strickland
Seattle, WA
On 5/7/07, Smith, Steven K - MSHA <Smith.Steven@xxxxxxx> wrote:
I'm researching data guard with physical and logical standby databases.
We currently are running standby database in 9.2 and are investigating
options that data guard will give us after upgrading to 10.2.
Question I have is - how reliable and what issues have people with
experience using logical standby? I see that there advantages would be
availability to report and view data with possibly additional reporting
indexes on the 'standby' server.. I understand that having that available
in physical mode is an option, but updates stop while the database is open
in 'read only' mode.
What is the performance hit? Assuming maximum performance mode. I don't
trust the network to the standby site to recommend Max Protection or Max
availability.
Ongoing maintenance? I know with our current standby database, the
maintenance on the standby site is minimal. The setup is just pretty
reliable as long as the network is available and not saturated.
Logical - large(r) bandwidth requirements?
I am currently reading the manuals so anything that I'll 'get to' please
don't tell me to RTM. I'm looking for more actual experience and lessons
learned.
Thanks
Steve Smith
Desk: 303-231-5499
Other related posts:
