row key formation in data_points column family of kairosdb

Question

I am new to Kairosdb, I am little confused about the rowkey formation in the data_points column family.

I am creating a metric with name "events" and it has two tags "sip" and "dip".

As per our use case, our application receives more than 10000 events per second, i.e. each event will have unique "sip" and "dip" fields. while I am writing data to this metric each unique combination would give unique row key in the data_points column family that means each unique entry would consume 1 partition of data_points.

Can any one please clarify is my understanding correct?

Since there is limit of 263 on the partitions(if murmr3partitioner is uesed) of data_points table. So I can not add more than 263 unique entries in the data_points table, i.e. across all metrics I can not add more than 2**63 unique entries.

Can any one please clarify is my understanding correct?

Answer 1

The hard limit is 2 Billions series per metric, and 2^63-1 partition (row keys).

2^63 is an amazingly high number (even substracted of one :), you should never come to that extreme level. cf. this answer: maximum number of partitions limit per table in cassandra