noticeboard.ru.ac.za

There's a known problem currently affecting some of our wireless access points. Users may find that they can successfully connect to a wireless access point, but that the won't be able to use the network from that access point. The access points affected by this vary from time to time, with no set pattern. It affects both the "Rhodes" and "campus.ru.ac.za" wireless networks.

This problem is indirectly linked to load shedding. We've got a work around for the problem, but each time the power is lost, the work around is undone.

We're currently looking for a more permanent solution.

All the wireless access points on campus connect back to one of two wireless security switches, one located in Struben building, and one located in the Africa Media Matrix. These switches provide the access points with their configuration, and ensure that wireless connections emerge on the right part of the University's network. In normal circumstances, an access point can chose to connect to either security switch and function in the same way. When they boot, they chose a switch at random for load balancing reasons.

At present any access point connecting to the Africa Media Matrix wireless security switch has been experiencing the problems described earlier. Because the security switch is selected at random, we've got no way to predict which access points this will affect at any given time.

Thus far we've worked around the problem by periodically rebooting the security switch in AMM. This causes all access points that were connecting to it to fail over to their redundant (and working) connection in Struben. Unfortunately, every time Eskom load sheds power they cause the access points to reboot, and to randomly select a security switch again.

We think that we have now managed to force all access points to chose the Struben wireless security switch by first preference when they boot. This should extend our work-around so that it survives Eskom's load shedding. We won't know, however, until the next power outage (i.e. Thursday).

Note that this doesn't yet fix the problem; it simply works around it so that users can continue using the wireless network. From a user perspective things should now "work", but there's currently no redundancy or resiliency on the wireless network. We'll continue trying to resolve the underlying problem.