startserver takes forever

[Audrey Won]

I am having trouble to bring up our ASE12.5.2 server. I am getting the log
message as below. It stayed at 'Started REDO pass..' for over an hour and
didn't seem to have any REDO done. I would like to restore the database from
a dump. At this point, is there a way to bypass the 'REDO pass'?
Thanks.

00:00000:00001:2008/02/03 18:20:41.33 server The optimal number of recovery
pro
cesses is 2.
00:00000:00014:2008/02/03 18:20:45.79 server Completed estimating recovery
log
boundaries for database 'facets_prd'.
00:00000:00014:2008/02/03 18:20:45.79 server Started ANALYSIS pass for
database
'facets_prd'.
02:00000:00014:2008/02/03 18:42:45.46 server Completed ANALYSIS pass for
databa
se 'facets_prd'.
02:00000:00014:2008/02/03 18:42:45.46 server Started REDO pass for database
'fa
cets_prd'. The total number of log records to process is 94481255.

[Manish Negandhi]

Post by Audrey Won
At this point, is there a way to bypass the 'REDO pass'?
00:00000:00001:2008/02/03 18:20:41.33 server  The optimal number of

recovery

Post by Audrey Won
pro> cesses is 2.
00:00000:00014:2008/02/03 18:20:45.79 server  Completed estimating recovery
log boundaries for database 'facets_prd'.
00:00000:00014:2008/02/03 18:20:45.79 server  Started ANALYSIS pass for
database  'facets_prd'.
02:00000:00014:2008/02/03 18:42:45.46 server  Completed ANALYSIS pass for
database 'facets_prd'.
02:00000:00014:2008/02/03 18:42:45.46 server  Started REDO pass for database
'facets_prd'. The total number of log records to process is 94481255.

High number of log records for REDO inidicates that there are many
committed transactions to be proceeded by recovery process.You can
bypass recovery of entire database not only REDO transactions. To make
sure your recovery process is not in hung state, you can connect to
the server while database is being recovered and look if value of
physical_io column of master...sysprocesses is changing

-HTH
Manish Negandhi
[TeamSybase Intern]

[Sherlock, Kevin]

You can't avoid recovery steps on startup unless you don't want the database
anymore.

ASE 12.5.2 is rather old software. I remember going through some recovery
performance pains in that release. 94 million log records is quite a lot.
You should probably use this opportunity to justify upgrading to at least
the latest 12.5.3 release (even better, go to the latest 12.5.4 release if
possible) which I believe includes some fixes to speed up some of these
recovery performance issues. I did that very thing on Christmas eve of
2005. It made all the difference in the world.

here is the bug fix info in the cover letter for the 12.5.3 IR (EBF 12143
for Solaris, CR 353375) :

12143 353375 The performance of the analysis and redo passes in
recovery for large numbers of transactions and
large numbers of log records has been improved.

Post by Audrey Won
I am having trouble to bring up our ASE12.5.2 server. I am getting the log
message as below. It stayed at 'Started REDO pass..' for over an hour and
didn't seem to have any REDO done. I would like to restore the database
from a dump. At this point, is there a way to bypass the 'REDO pass'?
Thanks.
00:00000:00001:2008/02/03 18:20:41.33 server The optimal number of
recovery pro
cesses is 2.
00:00000:00014:2008/02/03 18:20:45.79 server Completed estimating
recovery log
boundaries for database 'facets_prd'.
00:00000:00014:2008/02/03 18:20:45.79 server Started ANALYSIS pass for
database
'facets_prd'.
02:00000:00014:2008/02/03 18:42:45.46 server Completed ANALYSIS pass for
databa
se 'facets_prd'.
02:00000:00014:2008/02/03 18:42:45.46 server Started REDO pass for
database 'fa
cets_prd'. The total number of log records to process is 94481255.

[Audrey Won]

I don't need the database anymore. I can recreate it from scratch. Then what
is my option to bring up the server? Thanks.

Post by Sherlock, Kevin
You can't avoid recovery steps on startup unless you don't want the
database anymore.
ASE 12.5.2 is rather old software. I remember going through some recovery
performance pains in that release. 94 million log records is quite a lot.
You should probably use this opportunity to justify upgrading to at least
the latest 12.5.3 release (even better, go to the latest 12.5.4 release if
possible) which I believe includes some fixes to speed up some of these
recovery performance issues. I did that very thing on Christmas eve of
2005. It made all the difference in the world.
here is the bug fix info in the cover letter for the 12.5.3 IR (EBF 12143
12143 353375 The performance of the analysis and redo passes in
recovery for large numbers of transactions and
large numbers of log records has been improved.

Post by Audrey Won
I am having trouble to bring up our ASE12.5.2 server. I am getting the log
message as below. It stayed at 'Started REDO pass..' for over an hour and
didn't seem to have any REDO done. I would like to restore the database
from a dump. At this point, is there a way to bypass the 'REDO pass'?
Thanks.
00:00000:00001:2008/02/03 18:20:41.33 server The optimal number of
recovery pro
cesses is 2.
00:00000:00014:2008/02/03 18:20:45.79 server Completed estimating
recovery log
boundaries for database 'facets_prd'.
00:00000:00014:2008/02/03 18:20:45.79 server Started ANALYSIS pass for
database
'facets_prd'.
02:00000:00014:2008/02/03 18:42:45.46 server Completed ANALYSIS pass for
databa
se 'facets_prd'.
02:00000:00014:2008/02/03 18:42:45.46 server Started REDO pass for
database 'fa
cets_prd'. The total number of log records to process is 94481255.

[Bret Halford]

Post by Audrey Won
I don't need the database anymore. I can recreate it from scratch. Then what
is my option to bring up the server? Thanks.

You can boot the server with traceflag 3608 (put -T3608 as one of the
parameters
to dataserver in the RUN_SERVER file). This causes ASE to boot up and recover
only the master database. Drop the database you wish to drop, shutdown ASE,
remove
the traceflag, boot the server again.

-bret