Sounds like the loss of TCP connection is not detected quickly by the server.
Use a lower heartbeat value (e.g. 6-10 seconds): http://www.rabbitmq.com/heartbeats.html On Friday, 5 June 2015 06:04:04 UTC+3, Alexander Hudek wrote: > > I'm testing auto-recovery with a very simple scenario: > > 1. Make a single durable auto-delete queue. > 2. Attach a single subscriber to this queue that simple logs its message. > 3. Check that consumer count is 1. > 4. Cut the network between the client and server, then restore it. > > What happens here is that the client seems to recover and brings up a new > consumer thread on the test queue. However, if I check the consumer count > it now reads two consumers. The second, presumably original consumer > appears to eat messages, but never actually logs them. After waiting a very > long time (around 20 minutes maybe?) the ghost consumer disappears, but all > the messages it ate remain gone. > > What's going on here? Is there something I need to do to make the first > consumer die on connection failure? > > This is with 3.2. > > Alex > -- You received this message because you are subscribed to the Google Groups "clojure-rabbitmq" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure-rabbitmq+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
