[zeromq-dev] subscriber stopped receiving messages from XPUB socket

Winston Huang weiqingh at mail.com
Thu Mar 21 14:43:55 CET 2013


Thanks for your reply. I think I might have found the problem. I have a xpub socket that has about 10 subscribers. The following events happened:

1) one subscriber's hwm is reached and it's moved to the end of the pipes. (in zmq::dist_t::write).

2) another subscriber was terminated (zmq::xpub_t::xterminated), causing the pipe to be removed from the dist. however in zmq::dist_t::terminated, the terminated pipe was removed by moving the last pipe to the to-be-removed pipe's spot. therefore the deactivated pipe in step 1 is moved in the front of the pipes. in the meantime, the value of eligible and active are decremented. therefore the last eligible pipe (which was in front of the de-activated pipe before this event) now becomes in-eligible. and it will not receive any messages after this.

let me know if this makes any sense. it's hard for me to write a standalone test case like this. I hope my explanation is clear. And if you can suggest any fix, let me know. I can see one fix is to swap with the last eligible pipe and then delete that position.

----- Original Message -----
From: Pieter Hintjens
Sent: 03/15/13 09:15 AM
To: ZeroMQ development list
Subject: Re: [zeromq-dev] subscriber stopped receiving messages from XPUB socket

It sounds like a problem in the subscription forwarding, yet it's not clear how a subscriber could be affected by the publisher restarting, with the proxy in between. Do you need the proxy at all? First thing would be to connect subscribers directly to the publisher. If the problem then still happens we can try to make a reproducible test case. -Pieter On Fri, Mar 15, 2013 at 4:59 AM, Winston Huang <weiqingh at mail.com> wrote: > hi there, > > I have multiple (5-10) subscribers subscribing to the same topic published > by one publisher. They are connected via a XSUB-XPUB proxy. All the > subscribers are always up and the publisher may come and go at times. I have > noticed that at times, after the publisher is restarted, one of the > subscribers might stop receiving any messages at all. It's not a slow-joiner > kind of issue because the publisher continues to publish message every > second and that subscriber may not get any messages at all forever, whereas > other subscribers are getting messages at the same time. I also verified > that the subscriber is waiting for messages (it's calling the receive > function.) and if I restart the subscriber, it will get messages again. > > Could someone enlighten me what I may be doing wrong? Is there any thing I > should be looking into? > > Thanks in advance. > Winston > _______________________________________________ > zeromq-dev mailing list > zeromq-dev at lists.zeromq.org > http://lists.zeromq.org/mailman/listinfo/zeromq-dev > _______________________________________________ zeromq-dev mailing list zeromq-dev at lists.zeromq.org http://lists.zeromq.org/mailman/listinfo/zeromq-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20130321/48d5174a/attachment.htm>

More information about the zeromq-dev mailing list