Hi, It is an optimization, that runs operations which could be executed locally, in user thread. The same will work for cache operations too.
First of all, do you really need to send a message to all nodes, including the local? You can send the message to the remote nodes only, it will automatically resolve your problem. Additionally, you can create a different thread for such kind of operations and invoke sending from this pool. Evgenii ср, 21 нояб. 2018 г. в 03:06, Bob Newcomer <b...@badte.ch>: > Hello, would it be useful if I provided a test-case to help explain the > situation? > > > On Sat, Nov 17, 2018, at 11:27 AM, Bob Newcomer wrote: > > Hello, > > I have a distributed app that has an ignite client instance connected to a > grid. Every app registers for the same topic and uses it to broadcast > information to its peers and itself to perform async work in the ignite > thread pool. The handler for this topic may take several minutes to perform > its job, but I thought this would be OK because the docs for sendOrdered > say: > > Note that local listeners are always executed in public thread pool, no > matter default or withAsync() > <https://ignite.apache.org/releases/latest/javadoc/org/apache/ignite/IgniteMessaging.html#withAsync--> > mode is used. [1] > > [1] - > https://ignite.apache.org/releases/latest/javadoc/org/apache/ignite/IgniteMessaging.html#sendOrdered-java.lang.Object-java.lang.Object-long- > > However, in my testing when I send to the whole cluster, the message sent > to the local listener (the message topic handler registered in the same > ignite client instance as the one sending) gets handled in the same thread > as sendOrdered() and blocks the sendOrdered() until it has been completed. > > I had a look through the code and could not see an obvious way to make the > local handler happen on an ignite thread pool instead of blocking the > sender. How can I make the local handler get called in an ignite thread > pool? > > Appreciate any assistance. > > >
