On 3/14/07, xavi <[EMAIL PROTECTED]> wrote:
Hello again,
it seems that we might have found the problem. The ISP where we have our
ActiveMQ server hosted tells us that they have some protection system that
defines some inactivity timeouts at TCP session level. The timeout for the
port that we are using is set to 15 minutes, and this matches with the
problems we are seeing (we stop receiving messages after 15 minutes of
inactivity).
People from the ISP suggest us to send some KeepAlives periodically. It
seems that ActiveMQ can be configured to send these KeepAlives
automatically, but we haven't found how to do it for Stomp connections. We
tried this...
<transportConnector name="stomp"
uri="stomp://server:8000?transport.keepAlive=true"/>
and we've decreased the kernel's tcp KeepAlive time
(/proc/sys/net/ipv4/tcp_keepalive_time) to 60s, but the tcpdump doesn't show
any keepalive packets, so we guess there's something wrong in what we are
doing, but... what?
On the other hand, is there a way to set the KeepAlive time (not only
enabling it or disabling it) at ActiveMQ's level?
We've got support for keepalive & inactivity timeout in ActiveMQ
already; we use it for OpenWire. The problem with enabling this is it
would require the Stomp clients to send keep alive commands to keep
the connection active.
--
James
-------
http://radio.weblogs.com/0112098/
