Java Client: Connection pool recovery.
lic_gma at hotmail.com
Fri Jun 1 04:46:36 EDT 2012
We have a 4 nodes cluster and it works like a charm, our issue on
development (this is not happening in production but we are afraid it
might) is that when we are doing several concurrent store, if one thread
gives an exception (timeout getting connection from the pool or
connection refused), the thread seems to stay waiting forever for some
background tasks, so when the thread reaches the last line of code, it
never returns back, lets say, using a ThreadPool, so we basically end up
with our ThreadPool full with unfinished tasks:
More details, we are catching the Bucket and calling the operation
directly from it, so the bucket will use his respective Riak client.
More information about the riak-users