I’m doing a proof of concept with Influxdb + Grafana for a monitoring platform. After implementing it for a small data subset, when I loaded an entire day for analysis, influxdb takes a long time to execute the queries (high CPU consumption - starvation and excessive memory consumption - TSI in use).
The server has 4 CPUs and 16GB of memory.
My essential question should I have better performance that I have currently for about 3 million points / day?
SHOW SERIES CARDINALITY
Example of slowness:
This query is taking 24 seconds:
Jun 1 15:13:02 server1 influxd: ts=2020-06-01T14:13:02.022081Z lvl=info msg=“Executing query” log_id=0N82ObTW000 service=query query=“SELECT mean(total_duration) FROM performance.autogen.emas WHERE time >= 1590369394686ms AND time <= 1590370358767ms GROUP BY time(1m)”
Jun 1 15:13:26 server1 influxd: [httpd] ::1 - - [01/Jun/2020:15:13:02 +0100] "POST /query?chunked=true
It returns 17 records:
select count(*) from (SELECT mean(total_duration) FROM performance.autogen.emas WHERE time >= 1590369394686ms AND time <= 1590370358767ms GROUP BY time(1m));
This behaviour is common for all queries. From grafana is even worst - maybe because of fetch(?) and rendering charts?
Thanks in advance