Heyho,
Similar to the tryLock-issue discussed in another thread, I've a problem
with put().
Cache.put(...) can have a lot of reasons for failling. e.g.
java.lang.RuntimeException: org.infinispan.CacheException: Member
ISNode-35671 no longer in cluster....
javax.transaction.HeuristicMixedException (different reasons)
...
Reasons of failing puts are often normal cluster operations. e.g. new
node joins in, other one leaves ... Under load, this stuff causes lots
of transactions to fail.
For a single put was my solution just to try (limited times) again, when
exception appear.
For transactions this is a bit more complex, due I need to "replay" the
operations of the whole failed transaction.
Do you have any best practice for these situations? Makes the feature
request for a uncontended put makes sense?
--tf
Show replies by thread