Log Shipping from Primary Server to Backup and vise versa.

[shenanwei]

I have a primary server and backup server located in different physical
sites.
The primary server is live and ship logs to backup site every 5
minutes.
The primary server is being full online backed up every sunday locally.
Those logs are being applied to backup server every five minutes.
For example,
Primary site has power outage on tuesday, the last log was shipped was
S0001000.LOG. Backup server was brought to live, the first active log
was S0001001.LOG when started. Same time, the log was shipped back to
primary site. The backup site runs until sunday, the log till
S0002001.LOG was shipped to primary site.

Sunday, I need to stop the backup server and switch the primary server
up.
I think I can use the backup image on primary site of last sunday, and
apply all the log (from S0001001.LOG to S0002001.LOG)include those
generated by backup site, then bring it up.
Somebody has the similar experience?

[shenanwei]

I have a primary server and backup server located in different physical
sites, all running DB2 V8.1.5 on windows.
The primary server is live and ship logs to backup site every 5
minutes.
The primary server is being full online backed up every sunday locally.
Those logs are being applied to backup server every five minutes.
For example,
Primary site has power outage on tuesday, the last log was shipped was
S0001000.LOG. Backup server was brought to live, the first active log
was S0001001.LOG when started. Same time, the log was shipped back to
primary site. The backup site runs until sunday, the log till
S0002001.LOG was shipped to primary site.

Sunday, I need to stop the backup server and switch the primary server
up.
I think I can use the backup image on primary site of last sunday, and
apply all the log (from S0001001.LOG to S0002001.LOG)include those
generated by backup site, then bring it up.
Somebody has the similar experience?

[shenanwei]

I have a primary server and backup server located in different physical
sites, all running DB2 V8.1.5 on windows.
The primary server is live and ship logs to backup site every 5
minutes.
The primary server is being full online backed up every sunday locally.
Those logs are being applied to backup server every five minutes.
For example,
Primary site has power outage on tuesday, the last log was shipped was
S0001000.LOG. Backup server was brought to live, the first active log
was S0001001.LOG when started. Same time, the log was shipped back to
primary site. The backup site runs until sunday, the log till
S0002001.LOG was shipped to primary site.

Sunday, I need to stop the backup server and switch the primary server
up.
I think I can use the backup image on primary site of last sunday, and
apply all the log since then, there will be 2 parts
(from S0000800.LOG to S0001000.LOG on the primary site)
(from S0001001.LOG to S0002001.LOG on the backup site)
then bring it up.
Somebody has the similar experience?

[Phil Sherman]

Your proposed solution should NOT work. You ship logs every 5 minutes
which could accumulate up to five minutes of updates on the primary
server that are not shipped before it initially fails. Restoring logs
from the backup server, after it picked up the workload, could cause
difficulties.

The quickest way to move back to the primary server may be to stop
activity on the backup server, make an offline backup and restore it to
the primary server. You should use the quickest procedure; making a
backup and restoring or using yesterday's backup and logs. Yesterday's
backup with logs shipped to the backup (through S0001000) and logs
S0001001 through S0002001 should bring the primary server to the same
state as the backup server.

Phil Sherman
sh*******@gmail.com wrote:

I have a primary server and backup server located in different physical
sites.
The primary server is live and ship logs to backup site every 5
minutes.
The primary server is being full online backed up every sunday locally.
Those logs are being applied to backup server every five minutes.
For example,
Primary site has power outage on tuesday, the last log was shipped was
S0001000.LOG. Backup server was brought to live, the first active log
was S0001001.LOG when started. Same time, the log was shipped back to
primary site. The backup site runs until sunday, the log till
S0002001.LOG was shipped to primary site.

Sunday, I need to stop the backup server and switch the primary server
up.
I think I can use the backup image on primary site of last sunday, and
apply all the log (from S0001001.LOG to S0002001.LOG)include those
generated by backup site, then bring it up.
Somebody has the similar experience?

[shenanwei]

For the backup site is remote, it takes long time to bring the ***GB
backup image to primary site.
So I am wondering if I can use the image has been taken on primary site
of sunday, and apply the logs from both site. From primary site, I use
log flie since last backup till S0001000.LOG, but ignore all the rest
(for it crash), plus those generated on seconday site (from
S0001001.LOG to end)
I think there could be some loss because of some log was not shipped to
backup site (after S0001000.LOG), but it is acceptable here. I dont the
structure of logs, and hope they are consistent.

[Phil Sherman]

You may still run into a problem. When you terminated forward recovery
on the backup system at log S00001000, you would have lost any logged
updates that were not committed. (The commits would have been done
during the last <5 min of operation before the primary server crashed.)
When you try to restore on the primary server; you will have a
discontinuity in the logs where you start the new log from the backup
server.

I'd backup the database on the backup server immediately after
terminating rollforward and ship that backup to the primary server
immediately. It will be available to restore on the primary server and
you can start rollforward recovery on the primary server until you make
a clean shutdown on the backup server and ship the last log tape back to
the primary server. This way, the outage to return to the primary server
would be transmit and process of a single 5 min log.

Phil Sherman
sh*******@gmail.com wrote:

For the backup site is remote, it takes long time to bring the ***GB
backup image to primary site.
So I am wondering if I can use the image has been taken on primary site
of sunday, and apply the logs from both site. From primary site, I use
log flie since last backup till S0001000.LOG, but ignore all the rest
(for it crash), plus those generated on seconday site (from
S0001001.LOG to end)
I think there could be some loss because of some log was not shipped to
backup site (after S0001000.LOG), but it is acceptable here. I dont the
structure of logs, and hope they are consistent.