Mostly micro cavity chains , which are nothing but growing vacancy clusters  are observed along those grain boundaries subjected to the applied uniaxial tensile stresses, and  aligned  almost normal to the direction of the stress axis. Theoretically one can't image to have creep strain without having applied stress unless one deals with the relaxation stage that takes place after one  turns off the applied stress.  With the acception of   high temperature Herring-Nabarro creep,  there is a large amount  of dislocation climb  occurs due to the vacancy flow to the compression sides of  the dislocations having edge or mixed characters.  This  temperature dependent  process especially enhanced very much under the applied high stresses, and the creep rate shows almost  sin- hyperbolic dependence  on the stress. (Barrett & Nix). 

With regard to many different mechanisms of cavity nucleation, growth and coalescence (like accumulation of vacancies, dislocation pile-ups, grain boundary sliding) you may consult the book Fundamentals of creep in metals and alloys by Michael Kassner and Maria-Teresa Perez-Prado.

If  you think about modeling of creep damage in structures in the macroscopic sense, then the damage rate (damage is a process) is related to the equivalent creep rate (to capture growth of cavities) and the maximum positive principal stress (to reflect nucleation and coalescence processes). Examples of such damage evolution equations are presented in a series of papers by B. Dyson, D. Hayhurst and many others.

For Cr-Mo-steels used in fossil-fired power plants at temperatures above their critical temperature (e.g. above 673 - 723 K) the three stages of creep are encountered. Creep is due to the steady applied stress (e.g. pressure of 150 bars in superheater tubes, ...). The creep strain is the adaptation of the metal to the stress. The main mechanism is dislocation creep with a dynamical balance between production of dislocations and their annihilation by thermal effects during the secondary creep. Cavities of the order of 1 or a few angströms are formed (nucleation and growth). They migrate towards grain boundaries (mainly those perpendicular to the direction of main stress but not only) and triple points where they coalesce to form microvoids of the order 1 to a few micrometers diameter. With time going, chains of microvoids develop along the grain boundaries. They result in microcracks and ultimately in macroscopic cracks.       

If you wish to get the full rigorous treatment of this problem with applications using irreversible thermodynamics combined with the continuum mechanics dealing with thermal stress gradients,  please check these two attached papers if you have enough passions dealing with the mathematical expressions.

Best Regards