I think this is a problem of labeling.
The log recovery is a kind of a restart which log source is a log backup instead of
the log volume. From the point of view of the restart the restart failed - because it
was cancelled ('400'). I will think about it to avoid these confusing messages which
only want to tell that the restart (log recovery) was aborted and not that an error
was detected.
regards,
Uwe
>-----Original Message-----
>From: Alexei Novakov [mailto:[EMAIL PROTECTED]
>Sent: Friday, June 18, 2004 8:35 PM
>To: Hahn, Uwe; MaxDB mailing list.
>Subject: RE: SAPDB Standby database setup.
>
>
>Yes, here is the end of knldiag file:
>
>2004-06-16 13:42:06 2634 11565 startup DEVi
>started
>2004-06-16 13:42:06 2454 52101 RESTORE Filetype:
>file
>2004-06-16 13:42:06 2452 46 Log
>recovering log from tape from IOSeq 0 (MaxIOSeq from
>tape: -1)
>2004-06-16 13:42:06 2453 11000 vasynclo
>'/opt/project/var/database/backup/TEST1/log.backup.001'
>devno 84 T78
>2004-06-16 13:42:06 2375 12822 TASKING Thread
>2634 joining
>2004-06-16 13:42:06 2453 52024 RESTORE 56 pages
><- "ackup/TEST1/log.backup.001"
>2004-06-16 13:42:06 2454 52012 RESTORE new tape
>required 4300
>2004-06-16 13:42:06 2634 11566 stop DEVi
>stopped
>2004-06-16 13:42:14 2452 31 Log normal
>end of log found at off -1 lastseq 1598.
>2004-06-16 13:42:14 2452 49 Log
>last-redo-read#3622:TR2363(1)[EMAIL PROTECTED]'Commit':20040616:14606
>2004-06-16 13:42:14 2452 WNG 18 Log REDO:
>Aborted
>2004-06-16 13:42:14 2412 44 Log Savepoint
>requested by T37 reason 'Distance' (started).
>2004-06-16 13:42:14 2422 4 Pager SVP(1)
>Start Write Data
>2004-06-16 13:42:14 2422 5 Pager SVP(1)
>Stop Data IO, Pages: 7 IO: 6
>2004-06-16 13:42:14 2422 6 Pager SVP(2)
>Wait for last split, TaskId: 47
>2004-06-16 13:42:14 2422 7 Pager SVP(2)
>Stop Wait for last split, Pages: 0 IO: 0
>2004-06-16 13:42:14 2422 53070 SAVPOINT
>B20PREPARE_SVP: 6
>2004-06-16 13:42:14 2422 8 Pager SVP(3)
>Start Write Data
>2004-06-16 13:42:15 2422 9 Pager SVP(3)
>Stop Data IO, Pages: 1 IO: 1
>2004-06-16 13:42:15 2422 10 Pager SVP(3)
>Start Write Converter
>2004-06-16 13:42:15 2422 11 Pager SVP(3)
>Stop Converter IO, Pages: 8 IO: 8
>2004-06-16 13:42:15 2422 53071 SAVPOINT
>B20SVP_COMPLETED: 6
>2004-06-16 13:42:15 2412 ERR 52608 RESTART LOCAL:
>failed
>2004-06-16 13:42:15 2412 ERR 63 Log RESTART
>ERROR '400' => SHUTDOWN IS FORCED
>2004-06-16 13:42:15 2412 12696 DBSTATE Change
>DbState to 'SHUTDOWN'(24)
>--------------------------------------- current write
>position -----------------
>
>
>+++++++++++++++++++++++++++++++++++++++ Kernel Exit
>++++++++++++++++++++++++++++
>2004-06-16 13:42:15 0 12845 DBSTATE Kernel
>exited normal
>2004-06-16 13:42:15 0 12890 DIAGHIST Backup of
>diagnostic files will be forced at next restart
>2004-06-16 13:42:15 0 12808 DBSTATE Flushing
>knltrace pages
>2004-06-16 13:42:15 0 11560 COMMUNIC Releasing
> T79
>2004-06-16 13:42:15 0 12696 DBSTATE Change
>DbState to 'OFFLINE '(28)
>--------------------------------------- current write
>position -----------------
>
>
>It is clear that recovery was successful, but it is
>not clear to me what caused error and restart.
>
>Alexei.
>
>--- "Hahn, Uwe" <[EMAIL PROTECTED]> wrote:
>> Ok, 7.4.3.32 writes a savepoint if restore log is
>> cancelled.
>> If any error occurs during log redo and cancel is
>> treated like an error
>> then the kernel is switched to offline.
>>
>> If you execute recover_cancel after log recovery
>> then you should see some lines from a running
>> savepoint in the knldiag before kernel is switched
>> to offline.
>>
>> regards,
>> Uwe
>>
>> >-----Original Message-----
>> >From: Alexei Novakov
>> [mailto:[EMAIL PROTECTED]
>> >Sent: Thursday, June 17, 2004 7:12 PM
>> >To: Hahn, Uwe; MaxDB mailing list.
>> >Subject: RE: SAPDB Standby database setup.
>> >
>> >
>> >Hi Uwe,
>> >
>> >I am running SAPDB 7.4.3.32-1 (Value of
>> KERNELVERSION
>> >is KERNEL 7.4.3 BUILD 032-121-068-842) on Linux
>> SuSE
>> >8.2 (with all the latest patches). It is latest
>> 7.4.3
>> >build I guess. Also on recover_cancel the instance
>> >status becomes offline - is it supposed to be this
>> >way? Are there any prerequisites for this savepoint
>> to
>> >work, or should I activate it somehow?
>> >
>> >Regards.
>> >Alexei.
>> >
>> >--- "Hahn, Uwe" <[EMAIL PROTECTED]> wrote:
>> >> Hello Alexei,
>> >>
>> >> after recover_cancel (for log recovery) should a
>> >> savepoint be written.
>> >> This savepoint makes it possible to start log
>> >> recovery again at this point.
>> >>
>> >> But there was an error in some versions which
>> >> prevented such savepoint.
>> >> Could you give me your used release and build
>> >> please?
>> >> It seems to me that your using a version which
>> >> contains this error.
>> >>
>> >> regards,
>> >> Uwe
>> >>
>> >> >-----Original Message-----
>> >> >From: Alexei Novakov
>> >> [mailto:[EMAIL PROTECTED]
>> >> >Sent: Thursday, June 17, 2004 8:38 AM
>> >> >To: MaxDB mailing list.
>> >> >Subject: RE: SAPDB Standby database setup.
>> >> >
>> >> >
>> >> >Hi Tilo,
>> >> >
>> >> >Thanks for your response.
>> >> >
>> >> >> >1) Are there alternatives to load log backups
>> on
>> >> >> >standby instance other then do
>> >> >> >recover_start-recover_replace in one utility
>> >> >> session?
>> >> >>
>> >> >> End your log recovery with a recover_cancel
>> (DBM
>> >> >> Server should do that by default, when ending
>> >> with a
>> >> >> still open utility session.) And restore the
>> next
>> >> >> needed log backup with recover_start. You can
>> >> >> determine the next needed log backup with the
>> >> dbmcli
>> >> >> command db_restartinfo on the standby database
>> >> and a
>> >> >> look into the backup history of your source
>> >> >> database.
>> >> >
>> >> >Yes I tried this, but it seems that next time I
>> >> have
>> >> >to start recovery starting from the first log
>> >> backup
>> >> >anyways even though it was recovered in previous
>> >> >session already. I think it kills the idea of
>> >> feeding
>> >> >standby instance with sequential log backups. Am
>> I
>> >> >missing something?
>> >> >
>> >> >> >2) Can I extend one utility-backup-recover
>> >> session
>> >> >> >over multiple dbmcli calls?
>> >> >>
>> >> >> No yet. Most probably in the future.
>> >> >
>> >> >It would definitely be the solution for previous
>> >> >problem.
>> >> >
>> >> >> >3) Are there any other common solutions to
>> setup
>> >> >> >standby instance? I read article of Dr. Yves
>> >> >> Trudeau
>> >> >> >on the topic, but that solution is not
>> possible
>> >> in
>> >> >> our
>> >> >> >environment and since this article was
>> written 3
>> >> >> years
>> >> >> >passed, so I would expect that some
>> developments
>> >> >> has
>> >> >> >been done in this field.
>> >> >>
>> >> >> You could buy expensive hardware and implement
>> >> our
>> >> >> new Hot Standby solution, where the kernel of
>> the
>> >> >> standby databases share the same log volume as
>> >> the
>> >> >> master database and read the log entries, as
>> they
>> >> >> are written. But most probably that solution
>> will
>> >> be
>> >> >> too expensive.
>> >> >>
>> >> >> Best Regards,
>> >> >> Tilo Heinrich
>> >> >> SAP Labs Berlin
>> >> >
>> >> >Best regards.
>> >> >Alexei.
>> >> >
>> >> >
>> >> >
>> >> >__________________________________
>> >> >Do you Yahoo!?
>> >> >Yahoo! Mail is new and improved - Check it out!
>> >> >http://promotions.yahoo.com/new_mail
>> >> >
>> >> >--
>> >> >MaxDB Discussion Mailing List
>> >> >For list archives: http://lists.mysql.com/maxdb
>> >> >To unsubscribe:
>> >>
>> http://lists.mysql.com/[EMAIL PROTECTED]
>> >> >
>> >>
>> >
>> >
>> >
>> >
>> >__________________________________
>> >Do you Yahoo!?
>> >Yahoo! Mail Address AutoComplete - You start. We
>> finish.
>> >http://promotions.yahoo.com/new_mail
>> >
>>
>
>
>
>
>__________________________________
>Do you Yahoo!?
>Yahoo! Mail Address AutoComplete - You start. We finish.
>http://promotions.yahoo.com/new_mail
>
--
MaxDB Discussion Mailing List
For list archives: http://lists.mysql.com/maxdb
To unsubscribe: http://lists.mysql.com/[EMAIL PROTECTED]
