A systematized survey of the analytic results for retrival queues is given by discussing a generalized model A/B/s/m/O/H. Here, A and B describe the interarrival and service time distribution, respectively, s denotes the number of waiting positions in the queue, O is the capacity of an orbit for the retrying customers and \(H=\{H_ k\), \(k\geq 0\}\) denotes the loss model by which a customer joins to the orbit with probability \(1-H_ k\) after the k th unsuccessful retrial. The survey concentrates an single-server models of type M/G/i/m/O/H. For \(m=1\), \(O=\infty\), \(H=NL=\{H_ k=1\), \(k\geq 0\}\), the number N of customers in the system, the waiting time w, the number \(\eta\) of retrials, the length L of busy periods, the number M of customers served in such periods, the server idle time i and the distribution \(\pi\) of time intervals between consecutive departures are examined. For M/M/1/m/O/H with finite m, O and \(H=GL=\{H_ k=\alpha \leq 1\), \(k\geq 0\}\) the exact values of state probabilities p(i,j), i,j\(\geq 0\), are derived, where i and j are the numbers of customers in the waiting room and in the orbit respectively. With \(O=\infty\) and \(H=NL\) an iterative procedure to approximate p(i,j) is reported. In the case of batch arrivals the M/G/1 models are discussed. For the multi-server retrial queues only some major models, techniques dealing with the problems arising, and some main results are presented. Different full-availability and non-full-availability systems are reported. It is shown that the decomposition property, which is a common feature of queues with server vacations, is also true in retrial queues of the model M/G/s/m.