45 years data is good enough for Johansen.

As far as it is reasonably above 30. 

As far as it is reasonable above(a minimum of) 5  until 10 year

I agree with 30

i have always known of a series of nothing less than 30 with 45 series i believe it is ok.

It is not about number of years. The test is in a VAR so you need to think of degrees of freedom, not years. And it will anyway depend on the signal to noise ratio in your data. To know how far you are from asymptotics you need to set up a Monte Carlo and simulate calibrating the DGP to the variances in your data and data sample. Never listen to anybody who tells you x years is "enough": it will always depend on what you are studying!

Chiara Osbat is correct. From a practical point of view use all the data that is available. If there Is to much noise there is little that you can do. You should also take account of the speed that disequilibrium is likely dissipate. If this is slow you will need more data. I have seen simple cases where 20-30 years was sufficient but these are exceptional.

45 year data with 3 to 4 variables and lag length of 2 or less, is more than enough. Having a large sample size is not as important as having more degrees of freedom. You need to concentrate on the number of variables you take and the lag length you select.

There are not a minimun but I considere reasonable to have 20 o more years.

Good luck

I have 5 year monthly data (60 observation ) is that enough for johansen cointegration

It is most unlikely that you have enough data. See Chiara's and my earlier answer. There is no golden formula to determine the minimum number of observations. The length of time covered by your sample is more important than the number of observations. In VECM models one often finds that a disequilibrium has a half life of many years. (The half life is the time taken for a disequilibrium to be reduced by 1/2) A VECM analysis requires a multiple of that half life. If your system has very few variables, requires a small number of lags, has no seasonality and a disequilibrium with a half life of a few months you may have enough observations. This i doubt but you must judge this yourself.

45 years