2 Replies Latest reply: Jan 20, 2010 5:37 PM by oradebug RSS

    investigating Amazon Elastic Compute Cloud

    14419
      Greetings fellow Oracle professionals,

      I'm relatively new to Oracle Database administration. Currently I'm investigating the possibilities of setting-up a replication site in the Amazon EC2. From a business perspective this seems like a good idea. I'm having a hard time finding information on the operational perspective though. There seems to be very little documentation on performance issues and configuration best practice. What about scaling. I saw a nice graph about how it is almost linear, but I cant find close to not documentation on it. It seems to me that it involves quite a bit of work to scale up a database to a bigger instance. What about stuff like having the redolog files on the fastest discs behind the fastest controllers? How does that relate to an EBS? I' m thinking network-drive in stead of SCSI monster.

      Ok people, I would really appreciate you sharing your experiences and tips in this thread, pointers to technical documentation, performance test, stuff like that. It seems an aria the community hasn't invested in to much. Maybe someone knows why.

      Kind regards,

      Erik
        • 1. Re: investigating Amazon Elastic Compute Cloud
          14419
          Greetings fellow Oracle professionals,

          OK, I'm experimenting with the possibilities and will post the questions that I run into.

          Sofar, I've followed the tutorial setting-up a Large instance with EBS and E-IP. I've managed to get it running and logon to the database using sqlPlus and EM. As setting up this environment is only part of my responsibilities, I decided I wanted to make an AMI of the install, shut down the instance and start it back up when I'll continue working on it, thinking it would save me some money. In the AWS Management I could only terminate the image, not stop it. I guess its because there was no EBS Root device. I terminated the instance, and was left with, well, nothing really: a premature end of my first test....

          questions:

          1. Is it possible to install and configure Oracle, create an image, terminate the instance and create a new instance reflecting all the work done?
          2. Is it true I was unable to create an image because the AMI I used wasnt set up to run with a root EBS?
          3. Why are the oracle AMI's set up this way? Utilizing the on demand AWS features seem virtually impossible that way.
          4. Are the supplied Oracle Ami's beeing used in real life scenarios?
          5. What measures are begin taken to recover from an instance crash?

          Kind regards,

          Erik
          • 2. Re: investigating Amazon Elastic Compute Cloud
            oradebug
            erikros wrote:
            1. Is it possible to install and configure Oracle, create an image, terminate the instnce and create a new instance reflecting all the work done?
            Yes. You are meant to be storing the Oracle database files, and probably nearly everything else, including ORACLE_HOME, on EBS volumes. You can use a filesystem or ASM with these devices. If you build and store your database on the ephemeral storage that comes with the instance you will be left with nothing after you terminate the instance.
            2. Is it true I was unable to create an image because the AMI I used wasnt set up to run with a root EBS?
            It is true that you can only stop the instance if you are using Boot from EBS, but if you store your database files in an EBS volume, then you can terminate the instance, then later start a new one, re-attach the storage, mount the filesystems, and run the database. It is much easier if you store ORACLE_HOME in a volume, so that you do not have to make changes such as the parameter file and enabling ASM in a new instance's ORACLE_HOME. This is why it is advisable to use Boot from EBS, or one of the pivot_root AMIs to maintain the machine state throughout its lifetime.
            3. Why are the oracle AMI's set up this way? Utilizing the on demand AWS features seem virtually impossible that way.
            Not so! You just have to learn how to use the tool. EBS volumes are the piece you were missing to maintain persistence.
            4. Are the supplied Oracle Ami's beeing used in real life scenarios?
            I frequently use the Oracle-supplied AMIs for my students when I teach classes and for doing demos. I build the databases and (s)pfiles on EBS filesystems. If I terminate the instance, all I need is to mount the filesystem, add a symlink to the (s)pfile, and an ortatab entry.

            For production uses, I start with the supplied AMI, but I bundle them onto an EBS volume, then use a pivot_root AMI to boot from the volume. I have my own Oracle installs that I keep as EBS snapshots. These are available to be build into a new volume on any of my instances when I need, say, Oracle 11g. I use runInstaller -clone to incorporate them into the oraInventory on particular systems.
            5. What measures are begin taken to recover from an instance crash?
            I think I covered most of this, but in essence, if a database is stored on EBS, it can recover from a crash just like any other database. If you are somehow forced to create a new instance, you can just follow the guidelines above.

            Regards,

            Jeremiah Wilton
            Blue Gecko, Inc.
            The Oracle Cloud Experts
            [http://www.bluegecko.net|http://www.bluegecko.net]