There was a discussion about that earlier.
If you’re having (for example) a satellite uplink between your RouterOS machine and the RADIUS server, there will be HIGH latency which you can’t do anything against (apart from setting the timeout values higher in RouterOS, which has been deliberately regarding the max value some versions ago…).
well there is an easy way to ignore radius client timeout..
install multiple radius against single DB…request from hotspot to radius will be UDP…whereas request between radius and DB is tcp…now you all know reliablity of TCP vs. UDP.
we are running a countrywide isp…with just one oracle+solaris DB…and more than 20 radius on each node communicating with that single db…n no authentication problems/timeouts..
Each one set the radius timeout as he wants, and I configure my hotspots with a hight timeout vale because I have several roaming partners who do not sometimes have a very good response in time. Moreover, for this ans other reasons a set a high timeout response in my freeradius servers. So it is not strange that I put a higher than 3000 ms in my MT devices.
Do you think if in the next version this bug can be mended?
I have hundred ofm MT devices and I think this last solucion about having several RADIUS entries is a botch-up. Increasing radius timeout value is cleaner.
This is the specific problem because there are topology differences. moreover, when an aacount actives the roaming service, it is necessary to add to the time to access to my radius, the time to reach the roaming partner radius. During this time, my MT has to wait for a response.
For this reason I need to increase the radius timeout.