When the cooling system (DC4200 and Rack) are set to 'auto' on most of their switches and the thermal-load on the cluster-room is low, the Cooling system shuts itself down, most important, it powers down the punps in the secondary cooling-circuit. What happens? - After the Racks are coold down sufficiently (<20°C) the LCP-Units close their valves. The Cooling system notifies, thar no water needs to be cycled any more an shuts down the secondary punps. Temperature in the LCP-Units rises again, but the sec. pumps won't come up again.
Everything but the sec. pumps is in 'auto'-mode (there is always water circulation on the sec. circuit): As temperatur in the Racks gets below the limit, the LCPs shut their valves. As flow gets smaller, the Cooling system itself shuts it's valves on the sec. Circuit and lowers the cooling performance of the outside hearexchangers. As a result temperature in the prim. circuit rieses. When the Rack-temperature climbs up over ther threshold, the LCP-valves open again, but as the prim. circuit cooling-capacity is limited, the secondary-circuit water-temperature rises quite fast. LCP-valves open to 100% and the prim. circuit performance is adjusted accordingly. Rack-temperature starts to lower quite fast again. As you can see, we have enterd a circle. In other words the system oscillates. - view the Cooling-system's logged values.
With both heat-exchangers (one pump on the prim. and sec. side each) running, compared to only one of the two, the oszillation frequency is much slower and the amplitude only 1 to 3K.