ETL for CCA... How to... DR or HA...?

[tony]

Hi Folks!

Okay - so I'm looking at the ETL Service and seeing that it is not DR/HA friendly... or is it?

AFAIK, ETL processing invokes Java to ETL ODS data (Agg, Trans, etc.) - and the configuration for a PRIMARY and BACKUP ETL Service appear to be exactly the same.  The problem I foresee is that any Java processes issued by the ETL Service PRIMARY before it failed over to BACKUP would still be running - and, since the BACKUP is a mirror of the PRIMARY, it too would try to run the same Java ETL processes, causing duplication/errors/trouble...

The expected HA/DR Solution builds from Genesys documents refer to duplicates of everything - except the ETL Service -- which is [b]not [/b] replicated.  So - should I deploy just [b]one [/b] ETL Service and ensure that it is "bomb-proof"...?!?

...or is it just me...?

Another way to put the question is; does anyone have any workable solutions to enable DR/HA on the CCA ETL Service on Solaris, for a 7.2x deployment?

Every little helps... thanks chaps!

TT

[cavagnaro]

Lets put on this way. Remember that the whole data comes from ODS and ETL just eat ODS data and throws an understandable report for your business needs. Does Genesys will continue working without ETL? yes. Does ETL impacts on normal operational way? not at all. Can you live without ETL for a day or two as long as ODS is still storing information? Yes.
ETL doesn't need HA as it is not an essential component for Genesys to keep working. Another thing could be that ODS fails, if it fails then nothing will collect information.

[tony]

Cav - these are exactly my thoughts, too!

...but...

With a DR scenario, the (business) intention is to have something which can be run in exactly the same way from either of 2 locations.  So, if the Primary Site fails... I need to find an automated way to switch to the Secondary site.  My big concern is how to STOP any associated JAVA Processes from running. I can see problems if I simply STOP the Primary ETL Service and START the Secondary ETL Service, since there are a number of (Extraction, Transformation, Load, Aggregation, etc.) JAVA Processes "in progress" (but I am not sure if I am worrying uneccessarily..?) I think I need to find a way to gracefully exit those JAVA Processes, before starting the Secondary ETL Service - otherwise there may be exact duplicate JAVA Processes attempting the same ETL Tasks on the same ODS - even though there would be only one ETL Service running....

I need to draw up a "process" of switchover, for various scenarios;

If the ETL Service's processes fail and data is written to the emergency file. This could be invoked by an OS Alarm, monitoring the emergnecy file location (SNMP).
If the Site fails (DR). This would be obvious in SCI and could be automated through an Alarm Reaction.
If the ETL Service fails ("HA"). Again, this would be obvious in SCI and automated through an Alarm Reaction.

BUT this is [i]just [/i] the ETL Services - what about all of the spawned ETL JAVA Processes which are running?  I mentioned we are on Solaris - and I think I am looking for a way to ensure that the JAVA Processes exit naturally/gracefully, [i]before [/i] starting the Secondary ETL Service... Unless, in some way, the Secondary ETL  Service is "clever" and can detect which JAVA Processes are already running and simply waits/loops until the Primary ETL Service JAVA Processes exit naturally - and then takes over the Task automatically...?

I hope I explained that OK..?

Thanks cav!

TT