MQSeries.net :: View topic

crashdog · Posted: Mon Apr 20, 2020 1:29 pm Post subject:

Hello,
for educational purposes I've setup an RDQM installation using three virtual systems running on CentOS 7.6. I've installed MQ 9.1.4 (CD) according to the knowledge center's instruction (using the installer script InstallRDQMsupport). As a side note, I first tried to install this with the LTS version 9.1.0.0 (no patches used). It failed to install because of multiple unsolvable dependencies.
But here's the real question.
The VMs have it's own subnet for the queue managers in the 10.0.1.x space. My main Windows 10 (with MQ Client 9.1.4 CD) system does not have a nic in that address space but I created a static route on my switch.
Now, when I run a queue manager on the primary node everything works as expected (I can connect from Windows to the Qmgr). But when I set the preferred node to another system then the primary, the queue manager switches over as one would expect and the floating IP is attaching it self to the nic on the alternate node.
But trying to reach the queue manager on the switched node fails.
To be precise, I can ping the floating IP after the switch from Windows. Also the dns is resolvable. The listener is running on port 1414 and the port is opened on firewalld.
The queue manager is reachable from other Linux systems. Even from systems that do not have NICs in the same subnet as the queue managers. Which indicates that the static routing is basically working.
I've tested the connection with "openssl s_client -connect 10.0.1.14:1414" which works on the primary node, but not when switched (from Windows 10, but still works from the Linux system).

Any ideas what could cause this behavior ?

Switch command: