Is there a way to prevent data Overwrite

[biswaKL]

Is there a way to prevent data Overwrite?

If I am doing a push say 45 for certain timestamp with certain tags,

Then I push 12 for same timestamp and same tags, then the value will get over written with 12.
Is there a way to stop that and retain the first value only? So subsequest requests with same Timestamp and same tags should get rejected.

[brianhks]

At first I thought this wasn't possible, if kairos were to read data before each write it would slow things way down. But I think there is a way to get what you want.

When kairos writes a datapoint it specifies the timestamp of the write. This is kind of a historical thing why it does this. This timestamp determines what data overwrites what in Cassandra during a read and compaction. So it may be possible to run the timestamp backwards. Start with some fixed point in time and then calculate a number that always gets smaller. So new data will get inserted with an older write time than older data and cassandra will effectively throw it away.

I bet this can be done in the form of a plugin or configuration option to turn it on and off. And it won't adversely effect the speed.

[biswaKL]

Hi Brian, thats a nice idea. I can comeup with the logic to insert reverse timestamp,
But in the current API, we are not pushing the datawrite timestamp right? is it possible to send?

[brianhks]

This will require code changes to implement. The timestamp I'm referring to is not exposed through any api.

[biswaKL]

ohk, I will check the code when I have time.
For now my data goes through kafka queue, so I think i can query the DB for now.