Atomic data for electron impact ionization of all the elements from H to Ge are reviewed, the rates for these processes needing to be regularly updated following the publication of new experimental data and new theoretical calculations. Experimental cross sections, along with specific theoretical calculations when experimental data are missing, are fitted as functions of the electron energy, and from these fits ionization rate coefficients can be evaluated. It has been possible to take into account all elements but not all charge states of every element. Since the purpose of the paper is to update the ionization data evaluated and proposed in previous review papers, it is discussed if modifications are needed for the ions not considered. For highly ionized ions starting from the Ne-like iso-electronic sequence corrections do not appear necessary. On the other hand, except for Fe, for slightly ionized ions, specifically below the S-like iso-electronic sequence, the previously proposed data often underestimate the total ionization cross section, since only direct ionization channels have been considered and indirect processes have been neglected. Multiplicative correction coefficients are given to agree with recently published theoretical calculations. Experimental ionization data are considered, even when the presence of populated metastable levels (related to the electron density inside the source) is reported in the ion beams involved in the cross-section data measurements. We deem such a procedure acceptable when the proposed rates have to be included in codes that simulate the impurity behaviour in magnetic-confinement fusion devices, i.e., when radial transport is added to ionization and recombination to predict spatially resolved charge-state distributions. On the other hand, for astrophysical plasmas the contributions of metastable levels to the experimental data may represent a serious problem since, generally, the values of the electron densities that are involved are much lower than those in the ion sources. However, we critically investigated this problem and we found that the presence of metastables does not significantly modify the rates of most of the ions apart from a dozen. For this set of ions we provide different, corrected rates. Recombination is not considered since a review has been recently published.