A similar issue was reported recently in JIRA, but where the cause was long GCs.
I'll be adding a bit of safety to deal with these situations.
The reason this happens is cos the existing Java Hot Rod client is based around sync TCP
communications. A true async client would make this a lot easier. I have some ideas for it
for after Java 8 is integrated :)
Cheers,
[1]
https://issues.jboss.org/browse/ISPN-5221
On 2 Mar 2015, at 14:43, Stelios Koussouris
<stkousso(a)redhat.com> wrote:
Looking at Remote Listeners and Continuous Queries last week I noticed that after the
default 60secs of inactivity the socket maintained open for server to client push of grid
events was closed with java.net.SocketTimeoutException and the client had no way of
recovering.
I am not sure if this was intentional feature so that inactive connections are closed but
the way to overcome this was to set sockettimeout(0).
However, even this solution is not perfect as in an environment where client/server are
separated by a firewall the "inactive" connection could be closed by the
firewall.
It be useful to "refresh" the connection at SOCKET_TIMEOUT-x Time to ensure
this doesn't occur.
Stelios
_______________________________________________
infinispan-dev mailing list
infinispan-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/infinispan-dev
--
Galder Zamarreño
galder(a)redhat.com