05 February 2020

Distributed cache is a wide used technology that provides useful possibilities to share state whenever it necessary. Wildfly supports distributed cache through infinispan subsystem and actually it works well, but in case height load and concurrent data access you may run into a some issues like:

and others.

In my case i had two node cluster with next infinispan configuration:


distributed cache above means that number of copies are maintained, however this is typically less than the number of nodes in the cluster. From other point of view, to provide redundancy and fault tolerance you should configure enough amount of owners and obviously 2 is the necessary minimum here. So, in case usage small cluster and keep in mind the BUG, - i recommend use replicated-cache (all nodes in a cluster hold all keys)

Please, compare Which cache mode should I use? with your needs.


/profile=full-ha/subsystem=infinispan/cache-container=myCache/replicated-cache=userdata/component=transaction:add(mode=NON_DURABLE_XA, locking=OPTIMISTIC)

Note!, NON_DURABLE_XA doesn't keep any transaction recovery information and if you still getting Unable to acquire lock errors on application critical data - you can try to resolve it by some retry policy and fail-fast transaction:

/profile=full-ha/subsystem=infinispan/cache-container=myCache/distributed-cache=userdata/component=locking:write-attribute(name=acquire-timeout, value=0)