We describe the architecture and discuss our operational experience in running the off-line reconstruction farm of the CDFII experiment. The Linux PC-based farm performs a wide set of tasks, ranging from producing calibrations and primary event reconstruction to large scale ntuple production. The farm control software uses a standard Condor toolkit [The Condor project page, 〈http://www.cs.wisc.edu/condor〉] and the data handling part is based on SAM (Sequential Access via Metadata [Sequential data Access via Meta-data 〈http://d0db.fnal.gov/sam/〉]) software. During its lifetime, the CDFII experiment will integrate a large amount of data (several petabytes) and the data processing chain is one of the key components of the successful physics program of the experiment.