We present the first large sample of metal absorption systems in pairs of QSOs with sightlines separated by about 1 Mpc at z= 2. We found 690 absorption systems in the spectra of 310 QSOs in 170 pairings. Most systems show C iv or Mg ii absorption. When we see absorption in one QSO, the probability that we see absorption in the paired QSO, within about 500 km s−1, is at least ∼50 per cent at <100 kpc, declining rapidly to ∼8 per cent at 200–400 kpc, ∼0.8 per cent by 1–2 Mpc proper distance. Although we may occasionally see an individual absorbing halo in two sightlines, the absorber–absorber correlation is primarily a probe of the distribution of metals around galaxies and Mpc scale galaxy clustering. QSO absorption lines give redshifts errors of ∼23 km s−1, almost 10 times smaller than the error for galaxy spectra at these redshifts, hence we can measure clustering on small scales, around 0.5 Mpc proper, with a small sample. The distribution of 23 absorber–absorber coincidences separated by <2.5 Mpc at z∼ 2 is consistent with an origin in galaxies with a normal correlation function, normal systematic infalling velocities and low random pair‐wise velocity differences, more consistent with blue than with red galaxies. Absorption in gas flowing out from galaxies with a mean velocity of 250 km s−1 would produce more redshift elongation than we see. The fast winds detected by Adelberger et al. in the same ions account for less than 1/3 of the absorption systems we see. Such winds may be confined to the ultraviolet (UV) luminous star‐forming regions of Lyman break galaxies. If most galaxies have winds, they cannot extend to 40 kpc with large velocities, while continuing to make UV absorption that we can detect. This suggests that most metals seen in the intergalactic medium at z= 2 arrived long before. We see an excess of C iv absorbers, with an a posteriori probability of 0.0003, when a line of sight passes a foreground QSO. We see 16 absorbers where we expect 5.8 at 0–600 km s−1, on the front side of the partner QSO. At these velocities, we see an excess absorber in ∼6 per cent of sightlines that pass within 0.1–2.5 Mpc of a QSO, but in <2 per cent of cases when we look directly at a QSO. These transverse associated absorbers are not the normal line‐of‐sight associated absorbers that have a broader velocity distribution. Excluding the sightlines to us, the 3D distribution of 59 absorbers around 313 QSOs is approximately isotropic, except for the 1.5–2σ tendency for the excess C iv absorbers to be on the front side of the QSOs. Our QSO redshifts may be too large by ∼300 km s−1, or there might be a real asymmetry coming from a hypothetical anisotropy in the QSO UV emission, or from isotropic UV emission that lasted less than ∼1 Myr, possibilities suggested by the excess H i behind these QSOs. The velocity dispersion of the excess absorbers near the QSOs is small, ∼250 km s−1, suggesting that both these absorbers and the QSOs are in the blue sequence of galaxies. The probability of seeing absorption when a sig...