Hi Again Alan: I think something got confused along the way. While i do have a problem of my acct unique id being different depending on if the acct start comes with nas port of not, this behaviour is the same in all servers, and it is really not causing any issues other than the fact that i cannot really disconnect the user by sending and accounting packet. The solution for this is just changing the definition of acct-unique-id in all servers. The second problem that i have, which is unrelated with the above is the fact that even though i copy all accounting packets to the other two servers, they are mostlyin sync with server1, but not exactly. We were talking about this possibly being caused by either udp packets being lost or being sent/arriving out of order. Now, this is the part that i am trying to fix. Since there is little i can do about packets being reordered. and i do not see lost packets in my network, i was wondering if there is something else i can do to resolve this. The only thing that i can think of is trying to proxy over TCP, but i am definetely open not any suggestions. Thanks!. Regards On Jul 10, 2015 6:30 PM, "Alan DeKok" <aland@deployingradius.com> wrote:
On Jul 10, 2015, at 5:09 PM, Ricardo Larrañaga < ricardo.larranaga@gmail.com> wrote:
What do you mean by "making the system more robust"? I thought that using tcp would do that, unless you are talking about using master-master replication on the db?
No. I mean make the *system* more robust. Ensure that the RADIUS servers all behave exactly the same way. That they can handle packets with / without NAS-Port. That they all work correctly all of the time.
Since we were talking about the radius protocol being intrinsically error prone, y dont know if i understand what you mean by making the system more robust.
See your comments about NAS-Port. You said that the systems behaved differently depending on what they got from the NAS, and that this caused problems. So... make the systems robust to getting different data from the NAS.
It's about understanding what the systems do, and what the error cases are. Then, ensuring that the error cases are handled in a sane / consistent way, and making sure that the systems behaviour correctly, and consistently.
Alan DeKok.
- List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html