In the interest of full disclosure, I've heard of the following methodologies:
Advanced Replication
Threading export/import by schema
Quest SharePlex (in beta for Linux)
Are there any other enterprising ideas out there?
Printable View
In the interest of full disclosure, I've heard of the following methodologies:
Advanced Replication
Threading export/import by schema
Quest SharePlex (in beta for Linux)
Are there any other enterprising ideas out there?
2 Hrs in what sense ? Does your 2 Hrs include the installation of Oracle on Linux ? Does it include doing FTP of exported dump file or setting up Repliaction or the Shareplex?
Depends on what all are you counting in those 2 Hrs.
Amar,
2 hours total downtime.
2 hours from the time the HP database is shutdown to the time a connection on the same logical database can be made on a Linux server.
Perhaps I assumed too much on that post.
If you are talking about down time, then i think you don't even need a minute to do that :), cause an export or a replication need no down time, i don't have much idea about shareplex so no comments on that.
I'm not sure you understand my problem.
Migrating from HP to Linux requires an HP export and a Linux import.
A 300gig database would take a few days to export/copy dmp file/import.
That's downtime because I can accept no transactions during the process of export/copy dmp file/import.
Oh, and can you replicate cross-platform with Oracle Advanced Replication? (without setting up a bizillion snapshots)
Ok now the information is trickling down slowly.Quote:
Originally posted by KenEwald
I'm not sure you understand my problem.
Migrating from HP to Linux requires an HP export and a Linux import.
A 300gig database would take a few days to export/copy dmp file/import.
That's downtime because I can accept no transactions during the process of export/copy dmp file/import.
I don't think export would take days for a 300Gig DB. Only the import will.
Advanced replication is possible across platforms.
Advanced Replication is perty expensive ($20k/cpu).
How difficult is it to setup cross-platform (8.1.7)?
You're not talking about snapshots, are you?
Can tablespaces be transported between the two systems? Are they compatible in that respect, 'cos that would be your fastest way of doing it.
since when does advanced replication cost $20k/cpu? are you thinking about rac?Quote:
Originally posted by KenEwald
Advanced Replication is perty expensive ($20k/cpu).
How difficult is it to setup cross-platform (8.1.7)?
You're not talking about snapshots, are you?
Actually advanced replications comes FREE with enterprise edition - it is not one of the options that must be licensed sepparately.
Advanced replication is free.Quote:
Originally posted by KenEwald
Advanced Replication is perty expensive ($20k/cpu).
I am talking about multi-master Replication. But replicating 300Gig of data across DBlink is not a very good idea and that too for migration which would be a one time job.Quote:
Originally posted by KenEwald
How difficult is it to setup cross-platform (8.1.7)?
You're not talking about snapshots, are you?
I would prefer export/import (no other choice). Export would not take much time, may be few hours or a day, should not be more than that.
So once you have exported the full database (the only down time would be during the export) you can take you time to import it on the linux machine.
My strategy with export import would be
1. Set up the database on the Linux machine. Create all tablespaces before hand.
2. Schedule the migration for weekend when the users are less or nil.
3. Start the export in a restricted mode on friday night with the option FILESIZE = 10 GB or bigger. Depending upon what file size you want to FTP.
4. After the first Dump file is created, FTP the file(in compressed format) to linux machine, uncompress and start the import. (This way i do not have to wait for the whole export dump to be created. )
5. Disable the restricted session after the export is over.
6. Finish the import on the linux box.
Before this i would test one or two small table export and import from unix to the linux machine just to check any issues regarding the import export versions or the character set so that on the final day i don't waste time on these issues.
HTH
No.Quote:
Originally posted by slimdave
Can tablespaces be transported between the two systems? Are they compatible in that respect, 'cos that would be your fastest way of doing it.
You're missing the whole point. He want's the cut-over from one box to another to be no more than 2 hours. If I were in this situation, I wouldn't care if it took 6 months to setup, test, and sync all the data as long as my cut-over time was less than 2 hours.Quote:
Originally posted by adewri
Advanced replication is free.
I am talking about multi-master Replication. But replicating 300Gig of data across DBlink is not a very good idea and that too for migration which would be a one time job.
I would prefer export/import (no other choice). Export would not take much time, may be few hours or a day, should not be more than that.
So once you have exported the full database (the only down time would be during the export) you can take you time to import it on the linux machine.
My strategy with export import would be
1. Set up the database on the Linux machine. Create all tablespaces before hand.
2. Schedule the migration for weekend when the users are less or nil.
3. Start the export in a restricted mode on friday night with the option FILESIZE = 10 GB or bigger. Depending upon what file size you want to FTP.
4. After the first Dump file is created, FTP the file(in compressed format) to linux machine, uncompress and start the import. (This way i do not have to wait for the whole export dump to be created. )
5. Disable the restricted session after the export is over.
6. Finish the import on the linux box.
Before this i would test one or two small table export and import from unix to the linux machine just to check any issues regarding the import export versions or the character set so that on the final day i don't waste time on these issues.
HTH
Is a cut-over of 300Gig DB is possible in 2 Hrs.
1. Using Export. Could be possible, provided that my machine is so powerful that the DB gets exported in less than 2Hrs Flat.
2. Advanced replication. 300Gig transfer across a DBlink in 2 Hrs ???????? How about the 2GIG/s LAN card. :)
3. Shareplex ?????
Don't know, i think im confused, Ken can you clarify what you want?:rolleyes:
With Advanced replication you don't have to transfer 300GB in 2 hours.Quote:
Originally posted by adewri
2. Advanced replication. 300Gig transfer across a DBlink in 2 Hrs ???????? How about the 2GIG/s LAN card. :)
This can be a long slow process, everything you put in replication will stay in sync. This may be the best candidate for under 2h switch-over limit.
Except, I heard that Shareplex can do it online... could be wrong info.
AFAIK Shareplex can do it online and once the main data load is complete it reads changes directly from the redo logs it will have little impact on you primary system while it is being migrated.
However, Shareplex is quite expensive unless you can negotiate a limited use license for the migration task only.
Regards
I stand corrected. Yes, I was thinking of RAC.
Sounds like SharePlex is a viable option and that Advanced replication would work over a dblink.
How would I incrementally replicate? One schema at a time? Then how would I switch over to the new box. What residual mess would I have to clean up .. with the Linux box thinking it's a replicated server?
Thanks for your patience, I've got a million questions.
But there may be some issues with reading of Redologs via Shareplex.
Refer to
http://metalink.oracle.com/metalink/...T&p_id=97080.1
http://metalink.oracle.com/metalink/..._id=219628.995
Since you are in 8i i cannot suggest Oracle Streams which is what Shareplex does. Streams also read the redolog files and archivedlog files to capture the LCR (Logical Change Records) and put them in a queue, propagate to remote queue and then apply on the remote DB. Streams involves lot of technology like Adavanced Queuing, Logminer, Replication etc so it can get complicated while setting up, but once its done its clean, quick and no overhead.
Its similar to shareplex, only plus point is that since its an Oracle product, it can read the Redolog files with more accuracy (cos Oracle wont let anyone know it storage structure of the redolog files) than Shareplex.
HTH
Thanks for the heads up on that note, I really appreciate it.
Sounds like it's just not supported by Oracle and that reading data from redo is Quest's (support) problem.
Also, Oracle reserves the right to change it's redo file format at will. Something to keep in mind when upgrading if you're using SharePlex on an ongoing basis.
Thanks again