[infinispan-dev] Asymmetric clusters

[Sanne Grinovero]

Hello,

2010/9/29 Manik Surtani <manik at jboss.org>:
> This has been brought up a number of times - Paul re JBoss AS's use of
> Infinispan, Sanne re: Lucene and several others.

My problems are not related to Lucene, but blocks any use I'm
attempting with Infinispan if state-transfer is enabled and caches are
not started quick enough consistently on new nodes.
So in my case caches are started lazily according to service
initialization (several different services define new caches
programmatically, on demand).
I would be fine if statetransfer could ignore values relating to
unknown caches - but I assume this would lead to data loss on
distribution cases; because of the same issue I experience
ClassNotFoundExceptions, as I receive state about plugins which aren't
yet loaded and will be started using other classloaders.

Also distribution should be different per cache, to consider on the
hash wheel only nodes having the specific cache active.

> The problem is detailed in JIRA - https://jira.jboss.org/browse/ISPN-658 -
> along with the beginnings of a solution.  Have a look and let me know what
> you think (on this mail list, not JIRA).  My plan is to implement such
> support for 5.0.

Hopefully soon, this seems to me a blocker for anyone planning to be
elastic with Infinispan and start more than a single cache :/
Looking forward for suggestions for temporary workarounds, I'm going
to try using REPL only and ignore state-transfer related errors,
Mircea suggested me to look into interceptors to try this approach.

Cheers,
Sanne

> Cheers
> Manik
> --
> Manik Surtani
> manik at jboss.org
> Lead, Infinispan
> Lead, JBoss Cache
> http://www.infinispan.org
> http://www.jbosscache.org
>
>
>
>
> _______________________________________________
> infinispan-dev mailing list
> infinispan-dev at lists.jboss.org
> https://lists.jboss.org/mailman/listinfo/infinispan-dev
>