batchPoints with influxdb-java getting overwritten unless forced unique time?

My project is currently using influxdb-java to connect to Influx 1.8.2. Currently my code looks something like this:

InfluxDB connection = InfluxDBFactory.connect(server, client);
connection.enableBatch(100,10,TimeUnit.MILLISECONDS);
connection.setDatabase(database);
BatchPoints batchPoints = BatchPoints.database(database).build();

long currTime = System.currentTimeMillis()*1000000;
double[] data1 = getInfluxData();
for (int i=someInt; i < data1.length; i++){
    if (i % someInt == 0){
        double[] data2 = processData(data1, i-someInt, i);
        for (int j=0; j < data2.length; j++){
            Point p = Point.measurement(someTable)
                .time(currTime+i+j, TimeUnit.NANOSECONDS)   //Line A (j < someInt)
                .tag("someTag", "someTagValue")
                .addField("someField", data2[j])
                .build();
            batchPoints.point(p)
        }
}
connection.write(batchPoints);
connection.disableBatch();
connection.close();

Currently, Line A needs to be present, otherwise it seems my points overwrite themselves and only 1 point will actually get written out to someTable. Thus, Line A is a hacky solution but it's the only way I can get all points written out. The inspiration for the use of batchPoints like this came from influx's own example performance tests where they don't even specify a time. So what am I doing wrong that's causing the points to overwrite if not for Line A? Might be switching influxdb-client-java instead, just for context, in case that might alleviate this issue.

@Isaac -

If you are using InfluxDB 1.8+, I would consider switching to the influxdb-client-java as it seems more frequently maintained.

For your code sample using influxdb-java, the BatchPoints builder is not being used idiomatically. I'd say testWritePerformance in the example tests is also strange and that test is disabled.

I'm eliding much of your code.

// setup ...

BatchPoints.Builder batchPointsBuilder = BatchPoints.database(database);

// start for loops ...

batchPointsBuilder.point(p);

// finish loops ...

connection.write(batchPointsBuilder.build());


// setdown

BUT looking at the java code and yours, I don't think this is the problem as point(Point p) is appending Points to a List. So it shouldn't be overwriting here.

If you leave out the timestamp, all the points will get the same timestamp because the batch arrives all at once (this is how a left out timestamp is determined in InfluxDB - when the Point arrives). I can't see your data and tags but if all the points are identical with fields keys, tags keys, tag values, and timestamp, then they will overwrite/replace each other. I suspect your data/tags/fields are supposed to be different; double check it.

Lastly, I'd have you check (and share if you will) how you know that all the points are overwriting themselves? What is the query you are using?

~There's a typo in your sample code at for (int i=someInt; i < data.length; i++){ - should be data1.~

In InfluxDB all points having same tag values and timestamp even when having different field values are considered to be duplicate. So such data points are silently overwritten. This is as per the InfluxDB design.

Is it a limitation of influx that only one point can exist with the same tags and timestamp?

> Yes, this is by design.
>
> If a collision does occur, does the old one get overwritten?
>
> Yes, it is silently overwritten.
>
> This is documented but not well. I opened influxdata/influxdb.com#324
> to address that.