3 Replies Latest reply on Jun 15, 2018 9:36 AM by Nico Heinze

    Disaster Recovery Environment Configuration without HA

    siva krishna New Member

      Hi All ,

      Currently we have Informatica Production Environment which is already live and we are planning to have the DR set up now.

       

      https://network.informatica.com/message/13604#13604

      I am planning to have the VIirtual IP concept as mentioned by user190956 in the above discussion, but can someone please provide more  insight on this?

      below are my assumptions.

      1. VIP will be assigned to one DNS name which is different from PROD and DR host names. Below are the examples.

        like

      DNS Name: etlinfaprodcom / 10.10.10.6

      PROD Host Name: infaprod.com/10.10.10.4

      DR host Name: infadr.com / 10.10.10.5

       

      When we are running the Informatica on PROD host: Prod node will be updated with etlinfaprod.com and back end Network team will configure to point actual prod ip to VIP.

       

      etlinfaprod.com will point to 10.10.10.6 - which in turn point to PROD IP - 10.10.10.4

       

      During the failover Switch to DR, 10.10.10.6 will point to 10.10.10.5 .

      so at any point we will be having one VIP 10.10.10.6 on both PROD and DR

       

      However since our PROD is already live and it doesn't have VIP set up currently. It is directly pointing to host (infaprod.com). I will udpate that to etlinfaprod.com ).  But need few details on how to set up DR.

       

      1. How to set up DR now on DR host? Should we keep the same Domain and Repository Names including Oracle SID's?

      2. How the repository and Domains will be copied to DR. Can we have Data Guard Set up between PROD and DR? Or need to deploy the code on both PROD and DR simultaneously?

      3. Can we have SRDF replication for $PMRootDir File system from PROD to DR alone?

       

      I am looking for a solution without restoring domain and repository into DR node using Data Guard and SRDF/ any other File system copy.

       

      Any help in this regard is highly appreciated. Thanks in Advance!

       

      Thanks,

      --Siva Krishna

        • 1. Re: Disaster Recovery Environment Configuration without HA
          Nico Heinze Guru

          First one obvious misconception: the fact that you can have disaster recovery without HA via Virtual IP addresses means that you have two servers of which always only one is active (the other one is shut down); the virtual IP address points to the one node which is up and running.

          There is no need to "share" anything. The DB contents for all repositories (domain and application services) do not notice that the 10.10.10.6 address is not a "real" IP address; they always address the 10.10.10.6 address, and this will - thanks to the DNS settings - be implicitly rerouted to either 10.10.10.4 or 10.10.10.5, depending on which of the two machines is up and running. The whole Informatica domain does NOT know anything about running on either of two separate servers.

           

          Q1: domain, repository, and DB information must at all costs remain as they are (except that gateway info must point to etlinfaprod.com).

           

          Q2: you don't have to copy any DB contents; the only thing you have to copy is the file system tree from 10.10.10.4 to 10.10.10.5 at the same mount points with exactly the same directory structure.

          And of course it would make sense to move the whole .../server/infa_shared directory tree to a shared network mount; only then does DR work because only then all the files are accessible to whichever machine is up and running.

           

          Q3: what is SRDF?

           

          One word of warning: forget about this approach if you run on Windows. That won't work because the Windows registry contains keys which must exist on both servers. A disaster recovery can be set up in a Windows network as well, but that must be done differently.

           

          Final word: the hint I've written after Q2 is mandatory, not optional. You need a file system which either node can access, and that of course cannot be any local file system, it MUST be a shared file system. That's a plain fact, that's nothing to discuss in any way.

           

          Regards,

          Nico

          • 2. Re: Disaster Recovery Environment Configuration without HA
            siva krishna New Member

            Hi Nico,

            Thank you for your detailed explanation.

            Ours is Red Hat Enterprise Linux Server and Informatica 9.6.1 HF1

             

            Q1. Do we need Infa_shared directory as shared mount? If we keep entire PROD file system in DR, can't we run Informatica on DR? Or can we have all the File System in the shared mount. so that we don't need to copy anything to DR?

            Q2. We have suggested to go with SRDF replication and have node configured with DNS name. When PROD goes down, we can bring DR up and run without making any changes except DNS pointing to Actual DR IP?

            Q3. Since our PROD is already live, what kind of analysis is required before updating PROD node  with DNS name? If DNS pointing to prod IP, does it work if we update the node with the below command?

             

            <INFA_HOME>\isp\bin\infasetup.sh UpdateGatewayNode -na etlinfaprod:port -rst

             

             

            Thanks,

            --Siva

             

            • 3. Re: Disaster Recovery Environment Configuration without HA
              Nico Heinze Guru

              Hi Siva,

               

              Q1: Having the complete file system on a network share should do fine. Usually Informatica recommends having the installation (i.e. the binaries, shared libraries, and so on) on a local hard drive for performance reasons, but that's not a must-have. A shared network mount for everything is fine.

               

              Q2: I don't know SRDF, I only can say that if you work with virtual IP addresses it's indeed enough to switch the virtual IP address to point to the DR machine once PROD is down, that's correct.

               

              Q3: infasetup.sh UpdateGatewayNode is indeed what I had in mind, that should do fine.

               

              Regards,

              Nico