Given the abundance of goodness-of-fit heuristics that are out there, taking only the satisfying indicators would be cherry picking that leads to a highly biased "confirmation" of your model. Usually, depending on your field or on the outlet for your research, (on should hope that) there is an agreement about what indices you should report. Different indices indicate different properties of your model (e.g., is the covariance adequately modeled, is the model better than an independence model, is the fitness "good" in relation to the degrees of freedom, ...). In my opinion, it is best to choose two to three measures _before_ you have a look at the data and then report those. I suggest to look at CFI and RMSEA.

but in my problem, only Rmsea was given satisfied result. others were not given satified result. So i done EFA and again done CFA with more improved model. but also only RMSEA was satisfied