Yes we are doing sender.close() . Also why does the broker replay messages?
Nitin -----Original Message----- From: Gordon Sim [mailto:[email protected]] Sent: Tuesday, November 12, 2013 5:32 AM To: [email protected] Subject: Re: Problem/Explanation ? On 11/11/2013 08:38 PM, Nitin Shah wrote: > Hi, > > I wondered if someone on the Qpid mailing list can throw some light on this > issue we are seeing. > > We are using Qpid C++ brokers and clients of version 0.24. > > We are using Topic Exchanges. > > We create a sender on a particular session, then a message gets sent on that > sender followed by deleting the sender. What exactly do you mean by 'deleting' in this case? Are you calling Sender::close() for each of these? Are you waiting for the message sent asynchronously to complete before 'deleting' the sender? If you are sending a single message per sender then perhaps it would be better to call send(msg, true), which blocks until the broker has confirmed the message. > This is done a lot of times in a period of time. > What we are seeing is that the memory usage on Qpid is increasing and > a lot of memory allocations sitting on the std::deque and that number > increases over time > > There are 2 problems we are seeing: > > 1. That memory seems to be increasing on the std::deque list inside > Qpid. > > 2. When we fail over the broker, there seems to be a lot of messages > that get replayed on the topics to the clients looking for received messages > on the same topic. These are message that have already been received before. > > Below is a snipet of what we have seen. > > Note that in the second half below, there are allocations which show the > number of bytes@address for time in -Minutes:seconds. These are blocks > allocated that are "alive" so to speak. > > Can someone explain why we would be seeing this and perhaps how we can avoid > the "replays". --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
