I’m glad it was my code too, and expected it to be. At least now I have some additional ideas for debugging I’m the future.
On Mon, 9 Apr 2018 at 12:17 pm, Tim Bain <tb...@alumni.duke.edu> wrote: > I was hoping you'd say that, because if the broker was randomly failing to > deliver a couple of messages out of a huge number, that wasn't going to be > a lot of fun to track down. So I'm glad you figured it out, for both our > sakes. > > Tim > > On Sun, Apr 8, 2018, 6:58 PM Lionel van den Berg <lion...@gmail.com> > wrote: > >> Good news, it’s our bug. We have some logic that is automatically >> cleaning up inactive consumers, that logic was supposed to be turned off. >> >> >> On Mon, 9 Apr 2018 at 12:39 am, Tim Bain <tbai...@gmail.com> wrote: >> >>> The number of producers and the message volume should both be irrelevant. >>> Though of course if there's a bug, all bets are off. >>> >>> On Apr 8, 2018 5:53 AM, "Lionel van den Berg" <lion...@gmail.com> wrote: >>> >>> I’m going to try to get someone to try your suggestions at site tomorrow. >>> >>> Regarding your question, yes, the subscriber is a server that is started >>> before the producers and never restarted. Does it matter that there are >>> multiple producers to the topic. Also it’s a very low volume topic, there >>> is lots of data on other topics on the same connection. >>> >>> >>> >>> >>> On Sun, 8 Apr 2018 at 1:43 am, Tim Bain <tb...@alumni.duke.edu> wrote: >>> >>> > I'd suggest you use a JMX viewer such as JConsole, and navigate through >>> the >>> > org.apache.activemq MBeans to find the subscription in question. On the >>> > Attributes page under the subscription, you'll see counts for the >>> number >>> of >>> > messages enqueued and dequeued. >>> > >>> > If those don't match, it means your messages are not getting delivered >>> to >>> > the subscriber. (Could they be expiring before they can be consumed?) >>> > >>> > If they match, but are lower than the number you sent to the topic >>> while >>> > the consumer was subscribed, we'll have to dig deeper. >>> > >>> > Also, I understand from your response that these consumers have been >>> > subscribed and connected the entire time the messages were being sent. >>> Is >>> > that what you're saying? >>> > >>> > Tim >>> > >>> > On Fri, Apr 6, 2018 at 7:11 AM, Lionel van den Berg <lion...@gmail.com >>> > >>> > wrote: >>> > >>> > > Non-durable an using the same connection as other topics that are >>> still >>> > > working. No filters. >>> > > >>> > > >>> > > On Fri, 6 Apr 2018 at 11:03 pm, Tim Bain <tb...@alumni.duke.edu> >>> wrote: >>> > > >>> > > > Are the topic subscriptions durable or non-durable? If the latter, >>> were >>> > > the >>> > > > clients disconnected at the time the messages in question were >>> sent? >>> > > > >>> > > > Also, do the subscribers in question use any selectors? >>> > > > >>> > > > Tim >>> > > > >>> > > > On Fri, Apr 6, 2018, 12:26 AM Lionel van den Berg < >>> lion...@gmail.com> >>> > > > wrote: >>> > > > >>> > > > > Another kind of vague one. We have found from our application >>> logs >>> at >>> > > > site >>> > > > > that some messages that we are sending appear to be sent OK on >>> one >>> > end >>> > > > but >>> > > > > are never received by the subscriber. >>> > > > > >>> > > > > This seems to coincide with topics where the traffic volume is >>> low, >>> > > does >>> > > > > anyone have any pointers on where to look first? >>> > > > > >>> > > > >>> > > >>> > >>> >>