To make a long story short, DFT is not a convergent family of methods, therefore there is no ‘best’ DFT functional. Therefore, a validation is necessary to identify an acceptable computational set-up. Which type of validation depends on the properties that are meant to be investigated.
For example, if you are interested into the structural features of your compounds you may want to compare the geometry optimization of a model compound achieved with different combinations of basis sets and functional with the relevant structural data. But be aware that this setup, potentially matching your needs for the structural features of your compounds, may not be as well suited when frequency calculations are involved, or when absorption spectra are simulated.
I hope this may help you. In the scientific production of my group you may find several examples of similar studies, aiming at validating the setup for structural, spectroscopic or thermodynamic properties of coordination compounds.
If I should point to a decent starting setup, my advice would be the mPW1PW functional coupled with def2tzvp basis sets. But - as I wrote - this must be accurately verified and compared to different, and possibly more suitable, combinations of functional/basis sets.
Depends on how accurate you want certain interactions to be described. I would recommend you to take a look at GFN-xTB and GFN2-xTB as semiempirical methods extremely accurate for organometalic frameworks 😊👍