Given, I didn’t hear from anybody, I did my own test. I am trying to use Influxdb to support multiple tenants. So, earlier I was thinking of using measurement names for the tenants. I did multiple tests (at least 25 test runs) with various combinations of RAM and tenants (as measurement names and databases). I know that ID fields add to the cardinality, so in one of the run I had removed ID column from the tag list. However, as I need to run group-by on ID column, I didn’t have that choice. So, I brought ID column back to tagset.
Finally, at the end, for an Influxdb instance, I am settled with 32GB RAM, and one database per tenant. I was able to run that instance with 25 tenants (read 25 databases). Each of these 25 tenants has around 300K series, and I saw that influxdb memory consumption was hovering around a little over 50%.
I hope this helps somebody who is in similar situation.
1 Like
