MAC OS 10.12.1,python 2.7

I don't know whether you have got it resolved by now, and I just encountered the same keyerror issue today, for my situation it fixed by specifying the attribute "columns" explicitly with a string type '0' when create a DataFrame. As the code below:
df = pd.DataFrame(data=list(range(30)),index=pd.date_range(start='2014-11-16',periods=30, freq='H'),columns=['0'])

In fact, you can assign any string-based name to your columns.
If `columns=['something'] is suppressed, pandas assigns numerical id to them.

works:
df = pd.DataFrame(data=list(range(30)),index=pd.date_range(start='2014-11-16',periods=30, freq='H'),columns=['0'])

doesn't work:
df = pd.DataFrame(data=list(range(30)),index=pd.date_range(start='2014-11-16',periods=30, freq='H'),columns=[0])

works:
df = pd.DataFrame(data=list(range(30)),index=pd.date_range(start='2014-11-16',periods=30, freq='H'),columns=['your_column'])

doesn't work:
df = pd.DataFrame(data=list(range(30)),index=pd.date_range(start='2014-11-16',periods=30, freq='H'),columns=[99])

closing this since there's no issue

Actually, i fixed it.
And i think it's still need to take, because the module need more compatibility to pandas DataFrame.
I would have pull request after i take more testing about pandas and Mac OS X.
However, i think need more information about dose that suitable.
I fixed it by make the pandas DataFrame client convert the columns to string type for avoid potential key error.
I think the reason, make this problem be occurred, is pandas have automatically convert the key, when it's received the DataDrame.
If this assumption is wrong, inform me please.
On the other hand, inform me too, if the assumption is correct, but just for Mac OS X.
I would try to add some code to automatically identify whether the system is OS X.

Hi @SatoshiNewton ,
Cool ! that's is the approach I applied on my end to solve the bug... and I believe that is right way to solved it (or at least to make influxdb-client more robust when dealing with Pandas dataframes).