If the manuscripts relies "quite heavily" on references that you cannot read, you cannot meaningfully assess its contributions. I would suggest you decline to review it.
You could propose that the authors either identify possible reviewers that do speak Russian and/or Romanian, or that they submit to a more localized journal.
I diverge from JeffE's answer. I would agree with him if this were only a question of one or two (non-key) references. But this particular case really appears to me to be parallel to reviewing a paper in my field but not my specialty: if I personally get a paper on econometric forecasting but work in supply chain forecasting, then I will not have a good idea of the state of the art in this particular special subfield. So I can't assess whether the manuscript expands on this state of the art. In such a situation, I do not think it would be helpful to review the parts of the manuscript that I can assess and add a caveat - instead, as in the present question, I would decline reviewing.
After all, in either case it sounds like additional reviewers will be required. It would be better to bring this to the editor's attention as soon as possible, and not only after reviewing the paper myself.
(This suggests a third alternative: explain the situation to the editor and ask him what he would prefer.)