Recently on an Axapta Architecture course, we were informed that Axapta uses a very client-centered method for AOS selection (client sends out a UDP broadcast, which is answered by AOS’s with load information, and the client then selects an AOS). I have a number of questions with regards to this… - What happens when an AOS goes down or clients lose contact with their AOS for some reason? How do they determine that this connection has been lost (I assume they poll the server at some interval). Would all these clients then broadcast for a new AOS, do they wait a random amount of time… what exactly happens? - When AOS’s are specified in the config util (Client tab->Settings->Server host names), are these treated as recommended or enforced, ie. does the client directly contact only these AOS’s or does it still broadcast to all servers and give preference to those listed? Thanks, Matt
Hi, If AOS goes down, all clients connected to it go down also. Client uses only AOS:es spcified in the config util.
Even if there are alternate AOS’s available? That’s nuts! Isn’t one of the reasons for having multiple AOS’s to prevent clients from going down if one server falls over!? Thanks for the feedback.
Just an aside (sorry, didnt think of this until after the last post), if the AOS’s are clustered, and one of them goes down, would the clients not transparently continue without a hitch?
The cluster environment works like non-clustered AOS:es. All clients go down with the server. That’s why it is called load-balasing cluster.
Hi With AOS cluster: client sends out UDP package (Multicast) into network. If all machines are in the same subnet, you should only specify the cluster name in the client config. No AOS Name, No AOS IP Adresses, nothing. If the clients are not in the same subnet: only specify the IP adresses in the client config. Server managers on AOS machines will respond with number of users and the client will then decide to connect to which AOS.