This study investigates a microseismic swarm induced by injection operations in the Arkema-Vauvert salt field. The seismic activity in this field is monitored only by two permanent 3-component stations deployed in two wells. This study focuses on a period of 21 months (2004 January–2005 September) during which 1214 seismic events are located. The seismic activity is divided into three periods correlating with the water injection operations, highlighting a migration of the seismicity toward a thrust fault connecting the injection well and the production well. A waveform analysis reveals S-wave anisotropy, and focal mechanisms are computed using P, Sv and Sh amplitudes manually measured on anisotropy-corrected seismograms. First, synthetic resolution tests assess the reliability of the focal mechanisms determination from the two 3-component stations deployed in the field. Synthetic data are generated for 1056 earthquakes with various focal mechanisms and are perturbed with noise. The results indicate that the type of focal mechanism is correctly retrieved for 74 per cent of the synthetic earthquakes, but the uncertainties of the strike and rake are significant (from 15 to 45). Next, the focal mechanisms are computed for 532 real earthquakes. The solutions primarily correspond to a dip-slip/thrust fault type with subvertical NE–SW and subhorizontal N–S to NW–SE nodal planes. Correlations between the focal mechanisms and the spatio-temporal distribution of the seismic activity are noteworthy. The study shows it is possible to reliably retrieve double–couple focal mechanisms for some faulting geometries with two 3-component seismological stations. However, the reliability of the focal mechanism retrieval depends on the station configuration. Therefore, the addition of further stations would improve the results.