heartbeat not fully implemented? about erlzk HOT 10 CLOSED

https://github.com/huaban/erlzk/blob/master/src/erlzk_conn.erl#L203

I'm using the same logical like official java client did.

from erlzk.

I missed that bit - and the way handle-call timeouts are used to trigger the operation..

But still, if there are no messages coming back from the server nothing will happen. So in my test (where I don't drop the server, but just drop the packets) it won't reconnect.

I think this bit in ClientCnxn.java is what is missing :

     if (to <= 0) {
         throw new SessionTimeoutException(
                 "Client session timed out, have not heard from server in "
                         + clientCnxnSocket.getIdleRecv() + "ms"
                         + " for sessionid 0x"
                         + Long.toHexString(sessionId));
     }

.. this depends on a 'lastheard' timestamp which is updated every time something is received from the server - see calls to 'updateLastheard' in clientCnxnSocketNIO.doIO.

One way to do this could be to have a separate process (maybe an fsm) that acts as a timer which gets reset every time something is received from the server. If its not reset in time it can send a message back to erlzk_conn telling it to reconnect.

it would then do the same thing as this bit of code in erlzk_conn:

handle_info({tcp_closed, _Port}, State=#state{host=Host, port=Port, monitor=Monitor}) ->
    error_logger:error_msg("Connection to ~p:~p is broken, reconnect now~n", [Host, Port]),
    notify_monitor_server_state(Monitor, disconnected, Host, Port),

Anyway - please let me know if you think I've missed something. I can try and get this bit working and send the change back if you'd like.. thanks k

from erlzk.

You are right, and yes, any support is welcome.

from erlzk.

Cool. I'll get back to this over the next couple of weeks. I'd also like to implement KillSession - which I could do on a separate fork - I've developed a work-around for it in the meantime but it would be nice for it to be part of erlzk.
https://curator.apache.org/apidocs/org/apache/curator/test/KillSession.html

from erlzk.

Actually i just found this feature on my todo list of this project, close session then reconnect a new one if there is no package received from server in 2 / 3 * timeout.

And there are more on my list.

• readonly connection
• persistent watcher (after receive watched event just register a new watcher immediately)
• high-level api, like delete_all, create_all (like mkdir -p)
• recipe

from erlzk.

doh! I'll have a think about them ;-)

from erlzk.

Hi megayu - I have combined the two changes - KillSession and the heartbeat watcher - into my fork - allow a little more time for testing and I'll make a pull request for it soon. Thanks - k

from erlzk.

Pull request created - let me know how it looks to you. thanks - K

from erlzk.

#8

from erlzk.

Thanks for your work, and sorry for taking so long.

from erlzk.