I analyze records of the tsunami caused by the 2004 Sumatra -- Andaman M9.3 earthquake from tide gauges, seismic stations and satellites, in the central Indian Ocean. I demonstrate that the tide gauge records with sampling intervals of 4 minutes or more are seriously aliased and under-sample the tsunami waveform. Care needs to be taken when using this data for modeling this or future tsunami. The 1 minute sampling interval data accurately record the main features of the tsunami. The tsunami had significant energy at periods shorter than 1 minute and this resulted in, probably minor, aliasing of the 1 minute data. It is suggested that tide gauges intended for monitoring of tsunami use a sampling interval of at most 60 seconds, with effective anti aliasing filters being applied. Where the data is to be used for modeling at least 4 samples per minute data should be recorded.