Because of the way the replicator works, replication is always incremental. After the clients have been failed over, it's normal to run a "catch up" replication to make sure the most recent data has been migrated to the new grid.
As an alternative approach that will seed the new grid faster, you could specify the following flag during the initial replication to replicate only the most recent backup for each client:
--allsnapups=false
Once each client had one backup on the new grid, you could fail the clients over to it and they would run incrementals. After the clients are failed over, you would run a catch-up replication as above (without the allsnapups flag) to migrate the remaining data. You could also leave the old system in place until the last of the backups expire if that works better.
You can use the --allsnapups=false flag with both standard replication and root-to-root replication. Don't change or remove any flags (that would be really bad!), just add the --allsnapups=false flag to your replicate command. If you want to replicate the rest of the backups, just re-run the replicate without the --allsnapups=false flag.
no need to wait for first backup
It's generally a good idea to wait for at least one backup from each client to be replicated to the new system, otherwise any clients that have not yet completed replication will re-run an initial since their data will not yet be on the new system.
--fullcopy That recreates copy of the grid, and not replicating to /REPLICATED domain. What you saying is I should use --allsnapups=false for initial root replication (instead --fullcopy)? Whay would be command after that?
replicate --dstaddr=DST --dstid=root
--dstpassword=DST-ROOT-PASSWORD --id=root --ap=SRC-ROOT-PASSWORD? (without any extra parameters?)
In this case clients are failed over by just changing DNS entry (clients are pointing to the same DNS entry before and after failover) and no need to wait for first backup.
2. mcserver.sh --restore --restoretype=new-system. Restore system on destination.
3. Change DNS record to point to new grid. (do not switch off original grid for now).
4. Check that clients are backing themselves up.
5. nohup replicate --dstaddr=DST --dstid=root --dstpassword=DST-ROOT-PASSWORD --id=root --ap=SRC-ROOT-PASSWORD --fullcopy. Run "--fullcopy" replication AGAIN? That should be incremental? An will not interfere with step 2.
6. When replication is finished shutdown source grid.
ionthegeek
2 Intern
•
2K Posts
0
July 31st, 2012 07:00
That's correct.
ionthegeek
2 Intern
•
2K Posts
0
July 31st, 2012 06:00
Because of the way the replicator works, replication is always incremental. After the clients have been failed over, it's normal to run a "catch up" replication to make sure the most recent data has been migrated to the new grid.
As an alternative approach that will seed the new grid faster, you could specify the following flag during the initial replication to replicate only the most recent backup for each client:
--allsnapups=false
Once each client had one backup on the new grid, you could fail the clients over to it and they would run incrementals. After the clients are failed over, you would run a catch-up replication as above (without the allsnapups flag) to migrate the remaining data. You could also leave the old system in place until the last of the backups expire if that works better.
ionthegeek
2 Intern
•
2K Posts
0
July 31st, 2012 06:00
You can use the --allsnapups=false flag with both standard replication and root-to-root replication. Don't change or remove any flags (that would be really bad!), just add the --allsnapups=false flag to your replicate command. If you want to replicate the rest of the backups, just re-run the replicate without the --allsnapups=false flag.
It's generally a good idea to wait for at least one backup from each client to be replicated to the new system, otherwise any clients that have not yet completed replication will re-run an initial since their data will not yet be on the new system.
XRay1978
1 Rookie
•
61 Posts
0
July 31st, 2012 06:00
Hold on, sounds like you are not talking about root-to-root replication. I am using command
nohup replicate --dstaddr=DST --dstid=root
--dstpassword=DST-ROOT-PASSWORD --id=root --ap=SRC-ROOT-PASSWORD
--fullcopy That recreates copy of the grid, and not replicating to /REPLICATED domain. What you saying is I should use --allsnapups=false for initial root replication (instead --fullcopy)? Whay would be command after that?
replicate --dstaddr=DST --dstid=root
--dstpassword=DST-ROOT-PASSWORD --id=root --ap=SRC-ROOT-PASSWORD? (without any extra parameters?)
In this case clients are failed over by just changing DNS entry (clients are pointing to the same DNS entry before and after failover) and no need to wait for first backup.
XRay1978
1 Rookie
•
61 Posts
0
July 31st, 2012 07:00
So, just to confirm the steps (not using --allsnapups=false to simplify):
1. nohup replicate --dstaddr=DST --dstid=root --dstpassword=DST-ROOT-PASSWORD --id=root --ap=SRC-ROOT-PASSWORD --fullcopy. Runs initial replication.
2. mcserver.sh --restore --restoretype=new-system. Restore system on destination.
3. Change DNS record to point to new grid. (do not switch off original grid for now).
4. Check that clients are backing themselves up.
5. nohup replicate --dstaddr=DST --dstid=root --dstpassword=DST-ROOT-PASSWORD --id=root --ap=SRC-ROOT-PASSWORD --fullcopy. Run "--fullcopy" replication AGAIN? That should be incremental? An will not interfere with step 2.
6. When replication is finished shutdown source grid.
XRay1978
1 Rookie
•
61 Posts
0
July 31st, 2012 07:00
Will try that later. Starting source grid back.
XRay1978
1 Rookie
•
61 Posts
0
July 31st, 2012 08:00
Seems to be working! Great Thanks!