Various aspects of the effect of temperature on magnetic measurements at the observatories Paratunka (IKIR FEB RAS, Russia) and Choutuppal (CSIR-NGRI, India) are considered. Hardware details are described, including sets of external and built-in digital temperature sensors, and passive and active temperature control systems in measuring pavilions. Examples of seasonal and diurnal variations of the indoor and outdoor temperatures are given, showing that with the hardware temperature dependence declared by the developers of magnetometers, direct effects in the results of magnetic measurements can reach 1 nT or more. An effective way to reduce this temperature effect is to improve the temperature conditions in the pavilions with magnetometers. Thus, the use of additional passive thermal insulation of the pavilion at the Choutuppal Observatory made it possible to reduce daily variations in magnetic measurements by almost an order of magnitude. Additional indirect effects arise due to the temperature dependence of magnetic susceptibility, the high values of which are typical in Kamchatka due to surface rocks of volcanic origin. It is shown that real accounting of temperature effects, aimed at increasing the reliability of the obtained magnetic data, is a difficult task. Increasing the stability of the temperature conditions in which magnetic measurements are performed is seen as the most effective way to solve the problem.