@grant2 ok maybe you can advice on this
i found this Performance difference between Flux and InfluxQL · Issue #18088 · influxdata/influxdb · GitHub
Maybe you can give me some hint or maybe you have some insight (I do not expect solution but more hints).
We are using influx 1.8.10 and grafana 9.1.6. Now I am aware influx 1.8 is a bit old, but updating it to 2.0 requires data migraton which is a bit risky - we want to avoid doing it, since we have huge amount of data, and we are not very confident it will work smoothly
here is the dashboard:
https://testresults.qt.io/grafana/d/hb8ZpTH4k/flux_test_workitems_capacity?orgId=1
the dashboard should show integration system cpu capacity. The basic unit is workitem (equivalent to build atifact) we multiply duration by cpu used, and we try to group by by 1 hour if possible
The 2 time series panes are not exactly same - since influxQL and FLUX have different possiblities, but they try to achieve same goal - capacity of integration system expressed in cpu hours.
From grafana stats
data fetching took 11 secs 2 379 360 rows using influxQL
and 54 secs using Flux and 23 rows…
this is the schema for workitem:
results": [
{
“statement_id”: 0,
“series”: [
{
“name”: “workitem”,
“columns”: [
“time”,
“branch”,
“computeHost”,
“cpu_count”,
“duration”,
“host_arch”,
“host_compiler”,
“host_os”,
“host_os_version”,
“id”,
“path”,
“platform_id”,
“project”,
“sha1”,
“state”,
“target_arch”,
“target_compiler”,
“target_os”,
“target_os_version”,
“taskType”,
“type”
],
there is one thing that i know can slow us down (cpu was mistakenly stored as string) so we convert it - but its a bug that will be fixed in few days. But even ignoring it its still much slower than QL. Even if i only fetch workitems without checking details.