Do you have a test to reproduce the issue ?
On Thu, Oct 6, 2022 at 3:16 AM Jelmer Marinus <jelmer_mari...@hotmail.com> wrote: > Hi, > > During the testing of the message-counters (see other mail) we also seem > to have lost messages. It looks like these messages get lost during > redistribution or when they get produced to the cluster. Any thoughts on > that ? > > Some extra information (the same as for the message-counter test). > > We produced 5000 messages to a queue on each cluster node (3 node > symmetric cluster). When we consumed the messages through on of the cluster > nodes we didn't get all the messages. Sometimes we were missing 1/3 of the > messages (probably from one node). So when we produced 3 times 5000 > messages and we consumed them through one node, using redistribution to get > the messages from the other nodes, we sometimes only got 10000 messages of > the 15000 produced. The message-counters were back to zero and we were not > able to consumer more messages. There we no other consumers on the queue. > > Our set-up is as follows: > > * Artemis 2.25 running in a Docker container on Linux. > * Primary - Backup setup using replication (there was no replication > going on during the test) > * 3 node symmetric cluster (3 primary, 3 backup) > * One-to-one relationship between the addresses and the anycast queues. > * There was paging on the queues. > * ZGC garbage collector > > Best regards, > Jelmer > > -- Clebert Suconic
