Just having the residuals is not sufficient to diagnose the cause of such a problem.  Has anything changed from when it was converging at 27k iterations? What type of model is this? It looks like you are using the Reynolds Stress turbulence model. Is there are reason you are using this? Does it converge with the simple k-epsilon or SST models?

Look at the results. Do the results make sense? Where are the highest residuals (visualize the contours)? Are there bad volumes nearby? Look at the results when the residuals are better (27k) and compare them to what is happening when the residuals start to increase (30k). Has anything changed?

As you can see, there are numerous possibilities which cause this.