Recovery on UDB

[Raquel]

Hi!!

UDB manual states:

"DB2 reuses log names after restoring a database (with or without
rollforward recovery). When a rollforward operation completes
successfully, the last log that was used is truncated, and logging
begins with the next sequential log. Any log in the log path directory
with a sequence number greater than the last log used for rollforward
recovery is re-used."

So far so good. It then states "An archived log may be associated
with
two or more different log sequences for a database, because log file
names are reused". It further states that during a rollforward, "you
could continue to roll forward through either the top log sequence or
the (newer) bottom log sequence."

I don't understand this. If the logs are reused, how can an archived
log be associated with more than one log sequence? Suppose I have
archive logs 1 to 100. Then I restore and rollforward to archive log
53. This would mean that the archive logs 54 to 100 will be 'reused'.
So, how will archive logs 54 to 100 be associated with more than one
log sequence? I would think they would only be associated with one log
sequence...the latest one (the log sequence that will be written when
those archive logs are reused).

Also, while rollforwarding, what parameter do we code (in ROLLFORWARD
DATABASE statement) which decides whether we want to rollforward
through the "top log sequence or the (newer) bottom log sequence.".

What am I missing.

TIA
Raquel.

[P. Saint-Jacques]

See answers folded below. #######

Raquel wrote:

Hi!! .....clipped So far so good. It then states "An archived log may be associated
with
two or more different log sequences for a database, because log file
names are reused". It further states that during a rollforward, "you
could continue to roll forward through either the top log sequence or
the (newer) bottom log sequence."

I don't understand this. If the logs are reused, how can an archived
log be associated with more than one log sequence? Suppose I have
archive logs 1 to 100. Then I restore and rollforward to archive log
53. This would mean that the archive logs 54 to 100 will be 'reused'.
So, how will archive logs 54 to 100 be associated with more than one
log sequence? I would think they would only be associated with one log
sequence...the latest one (the log sequence that will be written when
those archive logs are reused). ################
Log files 1 to 100 in log path (and also archived) make up what we'll
call Log Chain #1
You restore and roll forward to Log #53. Let's presume that this point
in time is in the middle of the log. At the end of the RF, DB2 will
chop the log off at that point and start a new log chain. Log Chain #2.
If the old chain is stil in the log path, then thos logs (54-100) will
be reused and their current content will be wiped out. That's why it is
a new log chain.
The old Chain#1 in the archive is still valid to use to recover to now.
Make sure you copy and remove all logs of the new Cahin#2 out of the log
path before attempting to roll thru Chain#1.
Also, while rollforwarding, what parameter do we code (in ROLLFORWARD
DATABASE statement) which decides whether we want to rollforward
through the "top log sequence or the (newer) bottom log sequence.". ############### There's no parm to allow you to choose which Chain to
use. DB2 always looks to the current log path to find the logs it needs
to roll forward.
In your case, after your roll forward to a point in time, the logs in
the path are logs forming Log Chain#2 and that's what it will roll
forward thru.
If you needed to roll forward threu Chain#1; you would A) Copy out and
delete the current logs out the path, b) Copy in all logs forming Log
Chain#1 c) Restore, roll forward thru that.
What am I missing.

TIA
Raquel.

############## Now be extremely careful if you also use the userexit
for archiving (also thru even if you don't).

LOogs from Chain#1 are in the archive. From the moment you roll forward
thru a point in time, all new logs being archived will belong to Chain#2
and will overwrite any logs file of Chain #1 if they have the same number.

HTH, Pierre.
--
Pierre Saint-Jacques - Reply to: sesconsjunk at attglobaljunk dot com
Reconstruct address: Remove the two junk and replace at and dot by
their symbols.
IBM DB2 Cerified Solutions Expert - Administration
SES Consultants Inc.

[Fan Ruo Xin]

"P. Saint-Jacques" wrote:

Also, while rollforwarding, what parameter do we code (in ROLLFORWARD
DATABASE statement) which decides whether we want to rollforward
through the "top log sequence or the (newer) bottom log sequence.".

############### There's no parm to allow you to choose which Chain to
use. DB2 always looks to the current log path to find the logs it needs
to roll forward.

IIRC, You can use OVERFLOW LOG PATH option for rollforward command to apply the
archieved log files (top log sequence) from different log path, other than
current log path.

[P. Saint-Jacques]

Rosin, tell me I'm wrong. I thought that the roll forward would look
FIRST in the normal log path for logs ( and would finf the logs from
Chain2) before looking in the overflow log path. I think yours would
work if the normal default log path specified in the db cfg would be
empty!!!
Make sense or am I off the wall ?
Regards, Pierre.

Fan Ruo Xin wrote:

"P. Saint-Jacques" wrote:

Also, while rollforwarding, what parameter do we code (in ROLLFORWARD
DATABASE statement) which decides whether we want to rollforward
through the "top log sequence or the (newer) bottom log sequence.".

############### There's no parm to allow you to choose which Chain to
use. DB2 always looks to the current log path to find the logs it needs
to roll forward.

IIRC, You can use OVERFLOW LOG PATH option for rollforward command to apply the
archieved log files (top log sequence) from different log path, other than
current log path.

--
Pierre Saint-Jacques - Reply to: sesconsjunk at attglobaljunk dot com
Reconstruct address: Remove the two junk and replace at and dot by
their symbols.
IBM DB2 Cerified Solutions Expert - Administration
SES Consultants Inc.

[Fan Ruo Xin]

Hi, Pierre,
You are right! I got wrong.
Yes, we have to manually take care of removing those log files which left by the
bottom log sequence. Then specify the OVERFLOW LOG PATH to let the roll forward
process to forward through top log sequence.

Thanks,
FRX

"P. Saint-Jacques" wrote:

Rosin, tell me I'm wrong. I thought that the roll forward would look
FIRST in the normal log path for logs ( and would finf the logs from
Chain2) before looking in the overflow log path. I think yours would
work if the normal default log path specified in the db cfg would be
empty!!!
Make sense or am I off the wall ?
Regards, Pierre.

Fan Ruo Xin wrote:

"P. Saint-Jacques" wrote:

Also, while rollforwarding, what parameter do we code (in ROLLFORWARD
DATABASE statement) which decides whether we want to rollforward
through the "top log sequence or the (newer) bottom log sequence.".

############### There's no parm to allow you to choose which Chain to
use. DB2 always looks to the current log path to find the logs it needs
to roll forward.

IIRC, You can use OVERFLOW LOG PATH option for rollforward command to apply the
archieved log files (top log sequence) from different log path, other than
current log path.

--
Pierre Saint-Jacques - Reply to: sesconsjunk at attglobaljunk dot com
Reconstruct address: Remove the two junk and replace at and dot by
their symbols.
IBM DB2 Cerified Solutions Expert - Administration
SES Consultants Inc.

[Raquel]

Thanks for your response. It is much appreciated.

"The old Chain#1 in the archive is still valid to use to recover to
now". By archive, do you mean a path other than the "current log path
(SQLOGDIR)"?

Unless we 'manually' copy 'full' active logs (which I don't think
anyone does), the only way to copy the full active logs from "current log
path" to some other path (which you refer to as archive) is a USEREXIT.
Is it correct?

Also, in your mail you have suggestd:

"LOogs from Chain#1 are in the archive. From the moment you roll
forward thru a point in time, all new logs being archived will belong to
Chain#2 and will overwrite any logs file of Chain #1 if they have the same
number."

So, what is the way out? Do we 'rename' the archive logs (in Chain#1,
to some arbitrary name) before we start logging for Chain#2?

TIA
Raquel.