populated fields don't get unset upon update

[ctberthiaume]

edit: Moved original issue text to new issue #3 which describes a similar but slightly different missing data problem.

[vsivsi]

If a point is written with a field=value in one update, and then replaced with field=NA in the next, is the bad value removed from the DB? I think that's the case that we hit here last night...

[ctberthiaume]

I think the problem is that Influx doesn't actually write null or any equivalent in the NA case. To signify a field without a value you just don't include that field when you write the point. The problem is there's no way to erase the field for that point later. So the NA in the updated point can't replace the original numeric field value.

Somewhat related Influx issues and stack overflow pages (although these are more for dropping a whole column rather than a field in a single point):

• [Feature request] Drop field influxdata/influxdb#6150
• https://stackoverflow.com/questions/26767148/how-to-delete-columns-of-a-series-in-influxdb/
• Ensure no field value is null influxdata/influxdb#2429
• What exactly is the story on nulls? influxdata/docs.influxdata.com-ARCHIVE#717

[ctberthiaume]

I think the only answer currently supported by InfluxDB would be to drop the series before reimporting. In this case the series would be defined by measurement and cruise tag value.

[ctberthiaume]

Actually the potentially multiple series to drop would be all combinations of measurement + cruise + any other tags defined in the data set. So some initial query to discover the correct set of series to drop would need to be done, then drop those series, then import the file.