Correct EICR coding for a socket supplying a Server.

It is understandable (because of high protective conductor currents, unwanted tripping etc.) why someone would want to omit an RCD for a 32 A socket-outlet that is intended to supply a data/server rack.

Regulation 411.3.3 permits this, provided a risk assessment is in place. 

Guidance Note 5 (Section 8.2.3) provides some guidance of what sorts of mitigation ought to be in place where consideration is being given to undertaking a risk assessment to omit an RCD for a socket-outlet ... in this case, whether additional bonding is in place within the data/server rack, and for the data/server rack in the particular room.

gkenyon said:

omit an RCD for a 32 A socket-outlet that is intended to supply a data/server rack.

That is fine IF the data/server rack uses an UPS (Uninturupted Power Supply) but a lot don't have a UPS

Sergio Fernandez said:

That is fine IF the data/server rack uses an UPS (Uninturupted Power Supply) but a lot don't have a UPS

That's puzzled me - what's your thinking here? A UPS typically won't provide separation from the mains/earth so not really a substitute for shock protection precautions, nor are they much use against nuisance tripping as they'll only support the loads for a limited time (often only a few tens of minutes) - so likely still things go down before anyone can realise and get to the RCD to reset it (especially out of hours).

  - Andy.

Sergio Fernandez said:

That is fine IF the data/server rack uses an UPS (Uninturupted Power Supply) but a lot don't have a UPS

That's puzzled me - what's your thinking here? A UPS typically won't provide separation from the mains/earth so not really a substitute for shock protection precautions, nor are they much use against nuisance tripping as they'll only support the loads for a limited time (often only a few tens of minutes) - so likely still things go down before anyone can realise and get to the RCD to reset it (especially out of hours).

  - Andy.