Hi Ken,
429 | user | {call nextval('hibernate_sequence')}
|
| T | 365 | 381 | S | 21 |
0
This one belongs to an open transaction so it if a checkpoint pointed to
that entry (which is unlikely) it would not be played at recovery time
but will be played before the backend get disabled.
I then go to restore the backends from my backup through the console. I do
the following:
1. disable *
2. restore backend db_primary testbackup
3. restore backend db_backup testbackup
4. force checkpoint db_primary checkpoint_corresponding_to_the_backup (as
reported in "show dumps")
5. force checkpoint db_backup checkpoint_corresponding_to_the_backup (as
reported in "show dumps")
6. enable *
You should never use 'force checkpoint', the restore mechanism takes
care of setting the appropriate checkpoints corresponding to the dumps.
Using 'force checkpoint' you bypass all internal checks that could
detect if anything was wrong.
The result is:
=============
i) I don't see any of the operations from log_id's 429 - 432 written to my
backends
These were probably already played before the backend was actually disabled.
ii) I DO see the operations from log_id's 447 - 450 (inclusive) written to
my backends
Because the "hibernate_sequence" table sequence number is updated (by 1) as
a result of log_id 450 (I've confirmed this) then I believe that the
recovery process is having some sort of problem with log_id 430;
430 | user | insert into tablea (componentName, timestamp, id) values (?,
?, ?)
|
<!%S|abc|!%><!%L|1166491718006|!%><!%L|17|!%> | T |
I don't see what kind of problem you could have with such entry. Do you
have something like a 'duplicate key' error ?
You should try to use the standard backup/restore procedures described
in the documentation and avoid using expert mode commands.
Keep us posted with your progress,
Emmanuel
--
Emmanuel Cecchet
Chief Scientific Officer, Continuent
Blog: http://emanux.blogspot.com/
Open source: http://www.continuent.org
Corporate: http://www.continuent.com
Skype: emmanuel_cecchet
Cell: +33 687 342 685
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
