True High Availability Architecture

[Former Member]

Hello,

Is it truly possible to design a High Availability Business Objects architecture?

A clustered CMS, network FRS will reduce failure points. However if the machine that hosts IFRS/OFRS goes down, i would have to manually create/activate IFRS/OFRS in another cluster machine as we cannot have more than one FRS service running at a given time in the environment.

Is there a way to automate so that fail-over is automatic? Your feedback/advice is much appreciated.

Thanks

[Former Member]

Fail-over is automatic. So if you have I/O FRS on multiple machines pointing to the same root directory, you can have true high availability. The FRS services/daemons behave as active/passive. The first available FRS will be active and all other FRS services will remain passive unless the Active FRS becomes unavailable.

Load is not shared between FRS services.

[Former Member]

Thanks a lot guys.

So to summarize

If i have two nodes/machines in a cluster say A and B

I can run IFRS/OFRS on both these nodes at the same time as long as both point to a shared repository drive. In this case they would work in active/passive mode.

Let me know, if i have understood this correctly.

[former_member142702]

That is correct.

Regards,

Stratos

[alexander_schuchman]

When we ran two sets of IFRS and OFRS, we kept getting error messages in the BOE log for file locking.

We were told by SAP Support to ONLY run one set of IFRS/OFRS.

Anyone else have the same experience?

-Alex

[Former Member]

Hi Alex,

Considering the active/passive behavior described before

Did you see both your IFRS/OFRS'es to be "Enabled" and "Running".

What is their state when observed via CMC?

Thanks

[Former Member]

Alex,

Where is your FRS? Is it on NAS or SAN or Local Hard disk on the server?

[former_member142702]

You can have more than one FRS running in a cluster at once as long as they use the same shared directory. I admit that it does not realy make sense to have two FRS of the same kind running on a single node but you can start one instance per node and thus improve the failover capability of your system.

Regards,

Stratos