@Vincent-TP concerning the current firmware:

I've not tried updating to 1.3.3, as that specifically states that it's only updatable from 1.3.2 according to the firmware page:

The error message is unspecific:

I can access the corresponding AP, and the controller can, too:

I'm rather stumped what to do here. Thanks for getting back to me!

  @Vincent-TP this (i.e., custom upload) is how I convinced the system to upgrade to 1.3.2 (it offered me 1.3.4 out of the box). I'll try 1.3.3 now.

  @Vincent-TP yes, port 8043 is open for myself (that's how I made the screenshots of the Omada controller) and also open for the devices to connect to the controller. The controller can also connect to the APs, and vice versa, even though that has to happen through a firewall. Possibly, I've found the culprit: it seems that the Omada Controller does not have enough stack space and fails to serve the file. I've restarted the controller with -Xmx2048m, and now it seems that it upgrades fine to 1.3.2 for one device. I'll try the others.

  @Vincent-TP yes, the error message was equal for all tries. Generally, I don't think that this error message itself can be made clearer, as it is by principle correct (the APs WERE trying to access the controller to download the firmware, but/and the controller did NOT respond to the access in a way that served the firmware), and as this is a network access I of course thought that it was due to firewalling between controllers and APs (which the error message insinuates of course), which I could exclude and that's what made the error message so strange. ;-)

Generally, what would be really great is if the controller or a thread of the controller fails with a Java exception due to insufficient heap space, that the controller also posts this as a separate error message in the GUI, so that you can see it from the controller itself: increasing the heap space through control.sh is no problem (just adapt JAVA_ARGS), but if you don't know that you need to look in server.log for that message in this specific case (the controller worked for all the rest of its functionality), you'll not think about this as a possible error.

Thanks for pushing this to development, and if they need logs to see the Java exception, just get back to me!