Wednesday, 22 June 2016

How long does it take to migrate database (RDS) storage in AWS

We have a fairly small database in AWS using RDS.  This is an oracle database.  When getting ready for go-live we’ve migrated to provisioned IOPS to get some performance guarantees. 

image

You’ll seem from above that this took 13 hours! for 240GB.

IT would have been much quicker to recreate the database in the new storage area by restoring a snapshot to the database format that you want.  Remember that you are limited in how you can name the FQDN of the database instance – AWS kinda decides on this stuff on creation.  If you are using this – create a cname so that when you create your DB and get a new name – you can easily change all of the config in one fell swoop,

If you are doing all of this in cloudformation, you’ll need to parameterise and lookup the database FQDN when your machines are up.  i.e.

Your database might get a name like:  myJDE.ctfr998djdusfr.ap-southeast-2.rds.amazonaws.com:1521

You’ll need to pass this to the enterprise server host so that you can set tnsnames.ora (on all machines) with the correct value.  You can do this by passing user data into the template (so order of creation might be important).  some good examples of this are in https://s3.amazonaws.com/cloudformation-examples/BoostrappingApplicationsWithAWSCloudFormation.pdf

1 comment:

Shannon Moir said...

This depends on so many factors, it's hard for me to tell you. You need to consider the size of the machine, the storage, the IOPs and whether you are BYOL or not. Lots of questions, best thing to do is POC something.

Extending JDE to generative AI