A computer-implemented method includes identifying a source data object of a distributed computing environment. The distributed computing environment includes two or more storage nodes. The source data object exists as two or more slices. At least one of the slices is replicated on at least two storage nodes. The computer-implemented method further includes associating the source data object with a tape. The tape is written by a tape drive controlled from the distributed computing environment. The computer-implemented method further includes copying the source data object to the tape by, for each source slice of the two or more slices, in sequence: selecting a source node of the two or more storage nodes whereon the source slice is replicated, mounting the tape drive to the source node, appending the source slice to the tape, and unmounting the tape drive. A corresponding computer program product and computer system are also disclosed.