Re: GoldenGate Extract wants archivelog from before checkpoint
- From: Mark Van de Wiel <mark.van.de.wiel@xxxxxxxxx>
- To: dedba@xxxxxxxxxx
- Date: Thu, 12 Jul 2012 21:40:48 -0700
Tony,
There must have been an open transaction at the point where the extract
wants to start that began in your archive log sequence 187880, and
GoldenGate wants to read the full history of that transaction. Chances
are this is a session (e.g. Toad is pretty notorious for these kinds of
sessions) that opened a transaction, never did anything, and never will
do anything, but you don't know unless you capture the entire transaction.
The most recent version of GoldenGate addresses this challenge but since
it looks like you weren't running this upgrading now won't fix your problem.
There is a way to skip a single transaction and I believe this is
documented. However to be sure you only skip the one transaction you
cannot capture anymore you might want to talk to Oracle Support. Also,
of course if this isn't a completely idle transaction then you will miss
changes in your replication process. GoldenGate captures per transaction
and if the transaction is not in the list of transactions to track then
it ignores any DML/DDL for the transaction.
You can also reset the extract to start at a more recent point in time
but then you will likely skip more transactions, since GoldenGate will
only start capturing transactions that start after the point where you
reset it.
Hope this helps.
Mark.
On 07/12/2012 04:55 PM, De DBA wrote:
> G'day,
>
> I'm in the process of migrating a 9i database to 11g with the help of
> GoldenGate. Due to some misunderstanding somewhere, the GoldenGate
> extract on the 9i instance has been down for 2 weeks. The abend was
> caused be a relocation error, perhaps due to memory leaks? It is a very
> old non-supported version, after all. When we start it up now it abends
> with this message:
>
> 2012-07-12 18:45:13 GGS ERROR 190 Could not find archived log for
> sequence 187880 thread 1 under default destinations SQL<SELECT name FROM
> gv$archived_log WHERE sequence# = :ora_seq_no AND thread# =
> :ora_thread AND archived = 'YES' AND deleted = 'NO' AND
> name not like '+%' AN>, error retrieving redo file name for
> sequence 187880, archived = 1, use_alternate = 0.
> 2012-07-12 18:45:13 GGS ERROR 190 PROCESS ABENDING.
>
> Clear - it needs archived log sequence 187880 which was created (according to
> the database) on 27 June. That log is missing, not backed up and not
> retrievable. We are looking at starting all over again.. :(
>
> What is strange though is this line in the extract detailed info:
>
> Log Read Checkpoint Oracle Redo Logs
> 2012-06-30 10:00:00 Seqno 188166, RBA 189710336
>
>
> How now? We already are up to log sequence 188166 and still need 187880?
> Why? Am I missing something here? The command we use to restart the
> extract is
>
> GSSCI> start extract E1
>
> Should I give it a sequence number here? Any help will be appreciated.
>
> Version info:
>
> - Oracle EE 9.2.0.8
> - GoldenGate for Oracle 9i and HPUX B11.11 ( 10.4.0.19 )
> - HPUX 11.1
>
> Cheers,
> Tony
> --
> //www.freelists.org/webpage/oracle-l
>
>
--
//www.freelists.org/webpage/oracle-l
Other related posts:
