[jira] [Commented] (TC-303) Influx summary query returns results not found in the corresponding series query (i.e. max and min)

["Ryan Durfey (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/TC-303?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16048395#comment-16048395 ] 

Ryan Durfey commented on TC-303:
--------------------------------

Should influx.db be its own component?

> Influx summary query returns results not found in the corresponding series query (i.e. max and min) 
> ----------------------------------------------------------------------------------------------------
>
>                 Key: TC-303
>                 URL: https://issues.apache.org/jira/browse/TC-303
>             Project: Traffic Control
>          Issue Type: Bug
>          Components: Traffic Ops
>    Affects Versions: 2.1.0
>            Reporter: Dan Kirkwood
>            Priority: Minor
>              Labels: influx
>
> From https://github.com/Comcast/traffic_control/issues/539
> For example, the influx series query results may look like this:
> [
> [time, value],
> [time, 10],
> [time, 20],
> [time, 34],
> [time, 26]
> ]
> and the influx summary query results for the same timeframe may look like:
> {
> time: x,
> mean: y,
> min: 8,
> max: 75
> }
> notice how the min=8 and max=75 is strange because 8 and 75 are not found in the series query. this is because the series query is being grouped into 60s intervals (which is an average of 6 10s intervals) but the summary query looks at every value recorded in influx (on the 10s interval) between the timeframe.
> so in this example, i would expect min=10 and max=34
> maybe it's possible to run the summary query against the series query results?? like a subselect query like this:
> SELECT mean(value), percentile(value, 5), percentile(value, 95), percentile(value, 98), min(value), max(value), count(value) FROM (SELECT sum(value)/count(value) FROM tps_total WHERE cachegroup = 'total' AND deliveryservice = 'ds-name' AND time >='2015-09-17T03:38:00-06:00' AND time <= '2015-09-17T15:38:00-06:00' GROUP BY time(60s), cachegroup)
> this bug pertains to the following api endpoint /api/version/deliveryservice_stats.json when data source is influx.
> here are a couple sample influx queries:
> summary_query #-> $VAR1 = 'SELECT mean(value), percentile(value, 5), percentile(value, 95), percentile(value, 98), min(value), max(value), count(value) FROM tps_total WHERE time >= '2015-09-17T03:38:00-06:00' AND time <= '2015-09-17T15:38:00-06:00' AND cachegroup = 'total' AND deliveryservice = 'ds-name'';
> series_query #-> $VAR1 = 'SELECT sum(value)/count(value) FROM tps_total WHERE cachegroup = 'total' AND deliveryservice = 'ds-name' AND time >='2015-09-17T03:38:00-06:00' AND time <= '2015-09-17T15:38:00-06:00' GROUP BY time(60s), cachegroup';
> from [~mitchell852@apache.org]:
> this is not fixed. :( if you pass in another interval like 1h, this issue again occurs.
> this will require a summary query based on a series query or basically a nested query as the issue suggested and nested queries are not yet supported in influxdb - influxdb/influxdb#52



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)