hi folks

hibernate-3.0beta1.jar

is it correct to create a no-op second level cache provider that never puts anything in the cache and that always says an object is not present. like this i do not have to get involved with replicating caches over a cluster of WebSphere AppServers?

if i can't do this and i have to use a cluster-aware cache provider.....my issue is:
i have a piece of work executed within a transaction that proceeds to completion on the appserver without throwing an exception. the transaction control is delegated to websphere. subsequently the second level cache replication fails. does this rollback the peice of work that worked ok? why should a transaction rollback because a server that was never involved with it couldn't update its cache?

from using toplink, i learnt (maybe i am wrong with hibernate..?) that clustered cache is more trouble than it is worth. any experiences of using clustered caches with hibernate/ websphere?

any response much appreciated..am i missing something?

paul

Why not just not using the second level cache (default configuraton)?

hi and thanks

maybe i didn't explain what i wanted to say properly...

because i am using a clustered application server configuration,
i don't want any second level cache at all: i always want to hit the database. maybe i can set an expiry time of zero but this seems to be a cheat and anyway effort will be expended populating and evicting to/from the cache. so what i proposed was to write my own cache provider and use it. this cache will be very dumb and a request will never result in a cache hit because the implementation of the cache will always say the required object is absent from the cache.

i tried not specifying any second level cache in my hibernate conf but i got a noclassdeffound for ehcache which says to me this is the default cache
provider. i don't want any cache.

this is for a variety of reasons that includes, but is not limited to:
-for some very frequently used processes we use v complex pl/sql and for performance reasons, some of the CRUD of the application has to use stored procedures. second level caching means everyone writing to the db has to go via the cache, otherwise it will contain stale data. the massive majority of the operations is simple CRUD involving no more that 2 or 3 joins: a perfect match for ORM.

-i am not sure of the effectiveness of a clustered cache in the event that one of the participants fails. surely some code on startup after a failure will need to repopulate the server node with a fresh cache. equally, like i said, i do not want to rollback a transaction for the sole reason that one of the participating caches could not update itself.

hope this is clearer

thanks again

paul

Just don't put any <cache> tags in your mapping files, then no second-level caching will be used.

this doesn't work for me...

i have to reference to any <cache> tags in my mapping xml, however the hibernate startup says ehcache will be used

what am i doing wrong?

thanks

paul


[1/11/05 11:13:03:688 GMT] 20014a12 SettingsFacto I org.hibernate.cfg.SettingsFactory Automatic session close at end of transaction: disabled
[1/11/05 11:13:03:688 GMT] 20014a12 SettingsFacto I org.hibernate.cfg.SettingsFactory Cache provider: org.hibernate.cache.EhCacheProvider
[1/11/05 11:13:03:703 GMT] 20014a12 SettingsFacto I org.hibernate.cfg.SettingsFactory Second-level cache: enabled
[1/11/05 11:13:03:703 GMT] 20014a12 SettingsFacto I org.hibernate.cfg.SettingsFactory Optimize cache for minimal puts: disabled

here's my mapping doc

That doesn't mean its active for any persistent classes, it will always be turned on and "do nothing" however.