Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: Logical Standby Database Oracle 9.2 LGWR redo shipping

Re: Logical Standby Database Oracle 9.2 LGWR redo shipping

From: Mark Bole <makbo_at_pacbell.net>
Date: Thu, 01 Sep 2005 16:27:57 GMT
Message-ID: <h_FRe.4374$oJ2.1490@newssvr17.news.prodigy.com>


Replies interleaved below.

parag9i_at_gmail.com wrote:

> More details to your questions are
>
> 1. Oracle DB : 9.2 OS : AIX 5
> 2. My goal is to setup logical standby database on two different
> servers. Before that I want to setup standby db in test server to
> understand how it works, configure and test different scenarios.

Excellent -- testing should always be conducted first.

> Of,
> course when we talk about data guard it is for disaster recover only.

For disaster recovery, I would strongly recommend using a Data Guard physical standby instead. Logical standby has too many limitations (see documentation) and too many bugs (see Metalink) to truly be considered a rock-solid disaster recovery solution in version 9i (where it was first introduced).

> 3. I want to setup redo shipping thru LGWR not by ARCH process and for
> maximum availability

You might want to consider using Data Guard Broker as a front-end tool for configuring your Data Guard.

Going back to your original message, how did you determine that the changes aren't being committed in maximum availability mode? What useful information did you find in the trace files? From the docs: "If a LGWR destination fails, the destination automatically reverts to using the archiver (ARCn) process until the error is corrected." Also, "Redo logs are read and applied to a logical standby database when a log switch occurs, not as they arrive on the standby site."

Did you examine the DBA_LOGSTDBY_PROGRESS table? Are you sure that transactions are not being skipped as part of the logical standby configuration? Are you sure you have your TRANSACTION_CONSISTENCY setting (DBMS_LOGSTDBY package) such that "false" query results are not returned to SYS or SYSTEM users?

Again, these are all reasons why logical standby is not well-suited for front-line disaster recovery, especially in maximum availability / protection mode. There are too many ways for the logical standby to deviate from your primary. A physical standby, being an image copy, is much less likely to deviate.

> 5. I have not understood what you are talking about ls -alt ? On which
> directory you want to run this command ? I have already listed files by
> ls -ltr from bdump directory

Your -r option is overriding your -t option, try omitting it. I was simply pointing out that the newest files in any directory are usually the most interesting ones if you are an admin of any kind, YMMV.

[...]
> My logical standby is working fine, when there is switch in logfile.
> which I do not want this. I want LGWR to apply changes immediately if
> network is available. Kindly help me regarding this as I am doing this
> kind of data guard logical standby setup first time.

See quote from documentation above. Just because the change has been shipped to the standby on-line redo log does not necessarily mean it has been applied in the logical standby database (even though you still have maximum availability/protection in this scenario). I haven't tested maximum avail/protect mode with logical standby, so I can only go by the docs at this point.

Bottom line: in 9i at least, use a physical standby for disaster recovery, use a logical standby for other purposes such as reporting.

-Mark Bole Received on Thu Sep 01 2005 - 11:27:57 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US