While installing the node, you can choose either the server name as the name of the node or have a different name based on your company's rules and guidelines if defined already.. I think by default Informatica suggests a name for the node during the installation process. You have an option to go with the default or choose a different name for it. Informatica maintains these node names in the PO and PCSF tables during node creation and node election.
In a cluster of 3 servers will there be 3 node names?
There would be separate tables for storing grid information in the domain configuration database and all the nodes that grid consists of.
In case when we have to start services on DR due to issues on the prod site, how will that work? Should we create another grid for DR withe their respective nodes ? Will the services on this grid be on passive mode? Should the shared drive for this grid be using the one available in our DR site?
Yes you may create another Grid for DR to match the production environment. There are different ways to setup DR environment and shared file system is one of the ways to replicate data between 2 or more geographical sites. In your case, you can setup a grid using a shared file system on prod and have it shareable across the DR site. When the Prod is active, the shared file system (FS) on the DR server will be replicated based on a defined schedule and during this time the FS on DR will be in read-only mode, indicating the file system is inactive for users. The DR FS will be made active only when the replication is broken intentionally or unintentionally.
Yes, you have to create another GRID with DR nodes in the domain. Those DR nodes will not have services running until actual DR scenario (means those nodes are passive and not running in PROD domain) or when you want to perform DR test.
Regarding replication of storage data on DR system, You can have DR volumes replicated from PROD volumes by technique called Mirroring and when drives would in mirrored status they would be available only in read-only mode and not available on the Informatica DR nodes (i.e. no need to have them mounted on DR servers until required).
Now lets say you to perform DR test or in actual DR scenario, you will get this mirroring broken from storage team and then server team would mount these volumes on DR servers in read-write mode and then you will find all prod data on DR storage volumes as per the last replication. Now you can start Informatica services on DR nodes and you have actual DR running serving as PROD env.
And Ofcourse, Similarly you would also need to have Informatica domain and repository databases replicated.
We did a similar setup this is what we did.
Domain : Single Domain
Active Nodes 1,2
DR Nodes 3,4 (Join Existing Domain During Install)
Active Nodes and DR Nodes are in Single Grid - DR Nodes will be Down Always.
DB 1 - In Active Network (Primary)
DB 2 - In DR Network - DB2 Syncs with less than 30 Seconds delay in DR Network. (Secondary)
Infa_shared 1- In Active Network ( Mounted on Node1 , 2)
InfaShared 2 - in DR Network - Does Mirroring with less than a min Delay (Mounted on Node 3,4)
Start Domain on Node1 , 2 using DB1 and Infa_Shared 1 Storage
DR - Setup
Shutdown Node1 , 2 and DB 1, Infa_Shared 1 can stay Active
Start Domain on Node 3,4 and DB2 will act as Primary now and Since Infa_shared2 is mounted on 3 & 4 Nodes it uses the DR Storage.
Setup: In Domain You can configure Primary and Secondary DB in Connect String or You can use a Single TNS Alias and DNS Server should direct the connections to which ever DB is active. In this Scenario we need to ensure that DB 1 Is down so that incase DB1 is active the connections are not directed to DB1.
All 4 Nodes have to be gateway Nodes or minimum 1 Node in Active Network and 1 Node in DR Network has to be Gateway Nodes.
When You start the DR Site the Master Node will take over the Domain Ownership and Repository services.
Integration Services can be run on 4 Nodes GRID and Since the Node 1,2 Are down it will run on 3,4 only.
When the Active Network is ready, we used to do a reverse Sync of DB2 -> DB1 and InfaShared 2 -> InfaShared 1.
Shutdown Nodes 3,4 & DB2 after the Reverse Sync.
Start on Node 1, 2 and They will come up as Normal and wont see any difference.
Make Sure After Active Network is up the DB2 & InfaShared2 start Morroring for Next DR event.
Hope This Helps without maintaining multiple Domains. In this case you don't have to change assignments or make any changes. Just Shutdown Node 1,2 and Start Node3,4.
I have another question
With our current single node, we have some cronjobs which run some shell scripts under infa_share/Scripts internally using pmcmd commands.
When it goes to a multinode approach, where should these cronjobs be created?
How ever the jobs were created in Active Node same should be done on DR node only thing is it will be disabled.
My question is whether the cron job now needs to be created in all the nodes of the active cluster or should it be created on the network shared drive
If you are not creating them in shared drive you may want to create on all nodes. If they are created in Network shared you can use them from any node.
Great suggestions. All your inputs have been really helpful.
I have another question and do not want to create another thread for this question.
I have created a grid with 3 nodes. When I configure Domain in the client tools (Designer, Workflow Manager, Workflow Monitor or Repository Manager), what should be the node name given since my grid now has 3 nodes?
It is not mandate to configure all 3 nodes. How ever follow below rules.
a) Just provide the Domain Node(Master Node) name alone when configuring for faster detection Not Mandate.
b) Informatica services must be up & running on the node which is defined in Host(Node Name).
It automatically maintains the 3 Node info upon configuration and tries every node when master is down.
You need not configure the Domain on all client tools when you configure it in Repository manager or WorkFlow Manager you should be able to see in all tools, hope if you have the INFA_HOME variable in your system environment variables.
What will be the hostname for the admin console when we have 3 nodes in our cluster?