The way it's supposed to work is that all broker services are loaded before the network acceptors are activated so that no clients can actually connect until the broker is fully ready/available.
Also, any changes to runtime settings via the management API (e.g. to security-settings) should be replicated to the backup before the broker responds to the caller. Therefore, if the caller didn't get an error when they made their configuration change then they are assured that the backup has the change as well. > Is this behavior expected during a failover? No. > Could this be related to the broker not having fully loaded its configuration when the client attempts to create the consumer? Technically speaking that's possible, but it would be considered a bug and should be fixed. One thing that you might consider is that security is only checked when a consumer is first created which means any authorization-related changes occurring after the consumer was created won't be reflected until the consumer is created again (e.g. during failover). As always, a reproducible test-case is the best resource to assist any investigation. If you have one please pass it along. Justin On Tue, Jan 28, 2025 at 5:34 PM Havret <hav...@apache.org> wrote: > Hi, > > One of the users of the ArtemisNetClient library has reported an issue > during broker failover and wanted to check if this behavior is expected > before diving deeper into potential client-side issues. > > During failover, when the connection to the master node is lost and the > client connects to the slave node, the broker sometimes rejects the attempt > to create a consumer with the following exception: > AMQ119015: not authorized to create consumer. > > This suggests that the broker is not allowing the consumer to be created, > possibly because the configuration (e.g., security settings or other > parameters) has not been fully loaded or synchronized during the failover > process. > > - Is this behavior expected during a failover? > - Could this be related to the broker not having fully loaded its > configuration when the client attempts to create the consumer? > > Any insights or recommendations would be greatly appreciated to help > clarify whether this is expected behavior or something that needs further > investigation. > > Thanks in advance! > > Krzysztof >
