Oracle Coherence database sync while table refresh

Question

Hello,
I am using Oracle Coherence to cache a database table for read only puposes. However, my question is that this table needs to be fully refreshed every night and during this time I want to retain the cache in Oracle Coherence and do not want it to be recycled as well since it has to be available 24X7. Is there a way to do that? I am thinking of using a custom CacheStore class that will have its own local cache of the table or the other approach will be to look up in the shadow table while the primary table is being refreshed? Are there any other optimum ways to do this? The primary consideration for me is fast response.
Any help will be appreciated.
Thanks,
Jay

Answer 1

Hi Jay,
No, I am not suggesting to stop or start the JVMs. Everything will continue run as it was running ...
The Nightly refresh process that loads the data will load in non-active version of the cache (dataCache). Once the load is completed, it will update the MarkerCache (markerCache) with this version. All the client requests will follow something like this:
//Lookup in markerCache which version of the data cache is active
NamedCache markerCache=CacheFactory.getCache("markerCache");
String str=markerCache.get("dataCacheName");
//The cache which has the refreshed and updated data (dataCachev1 or dataCachev2)
NamedCache dataCache=CacheFactory.getCache(str);
For the nightly refresh process, you will do something like this:
NamedCache markerCache=CacheFactory.getCache("markerCache");
String str=markerCache.get("dataCacheName");
if ("dataCachev1".equals(str)){
       NamedCache newDataCache=CacheFactory.getCache("dataCachev2");
       //load the data{ }
       //update the marker Cache with this newly loaded cache; from now on the client will start using this newly loaded cache
       markerCache.get("dataCacheName","dataCachev2");
       // wait for few minutes to let all the old request complete
       //destroy the old cache
       CacheFactory.getCache("dataCachev1").destroy();
} else{
    //do the reverse for dataCachev1 and dataCachev2
}The only thing you need to note is that there might still be some requests using the dataCachev1 when it is destroyed so you need to handle the exception gracefully at the client application.
Please note, that you should optimize the code to minimize the time spent in lookup of the markerCache by using Maplisteners as appropriate.
Hope this helps!
Cheers,
NJ