[
https://issues.apache.org/jira/browse/TC-270?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15998445#comment-15998445
]
ASF GitHub Bot commented on TC-270:
-----------------------------------
Github user mitchell852 commented on a diff in the pull request:
https://github.com/apache/incubator-trafficcontrol/pull/545#discussion_r115017970
--- Diff:
traffic_ops/app/lib/Extensions/TrafficStats/Builder/CacheStatsBuilder.pm ---
@@ -74,14 +74,12 @@ sub summary_query {
if ( $self->validate_keys() ) {
#'summary' section
- my $query = sprintf(
- '%s "%s" %s',
- "SELECT mean(value), percentile(value, 5),
percentile(value, 95), percentile(value, 98), min(value), max(value),
sum(value), count(value) FROM",
- $args->{series_name}, "WHERE cdn = '$args->{cdn_name}'
AND
- time >
'$args->{start_date}' AND
- time < '$args->{end_date}'
- GROUP BY
time($args->{interval}), cdn"
- );
+ my $query = qq[SELECT mean(value), percentile(value, 5),
percentile(value, 95), percentile(value, 98), min(value), max(value),
sum(value), count(value)
+ FROM "monthly"."$args->{series_name}.cdn.1min"
--- End diff --
The RT name is also hard-coded in the deliveryservice_stats endpoint -
https://github.com/apache/incubator-trafficcontrol/blob/master/traffic_ops/app/lib/Extensions/TrafficStats/Builder/DeliveryServiceStatsBuilder.pm#L81
How about I create a new Jira and a subsequent PR that removes hardcoded RT
references from both cache_stats and deliveryservice_stats endpoints and maybe
puts the RT names in this file:
https://github.com/apache/incubator-trafficcontrol/blob/master/traffic_ops/app/conf/production/influxdb.conf
> GET api/cache_stats needs to query summarized influx data rather than raw data
> ------------------------------------------------------------------------------
>
> Key: TC-270
> URL: https://issues.apache.org/jira/browse/TC-270
> Project: Traffic Control
> Issue Type: Improvement
> Components: Traffic Ops, Traffic Stats
> Reporter: Jeremy Mitchell
> Assignee: Jeremy Mitchell
> Priority: Minor
> Fix For: 2.1.0
>
>
> The following API queries raw influx data causing it to be slow and thus it
> times out:
> GET api/1.2/cache_stats
> That API endpoint currently runs 2 queries like such:
> series query:
> SELECT sum(value)*1000/6 FROM "metricName" WHERE time >
> '2017-05-03T09:22:00-06:00' AND time < '2017-05-04T09:22:00-06:00' AND cdn =
> 'cdn1' GROUP BY time(60s), cdn ORDER BY asc
> summary query:
> SELECT mean(value), percentile(value, 5), percentile(value, 95),
> percentile(value, 98), min(value), max(value), sum(value), count(value) FROM
> "metricName" WHERE cdn = 'cdn1' AND time > '2017-05-03T09:22:00-06:00' AND
> time < '2017-05-04T09:22:00-06:00' GROUP BY time(60s), cdn
> Those queries should leverage the summarized data and look like this instead:
> series query:
> SELECT sum(value)/count(value) FROM "monthly"."metricName.cdn.1min" WHERE cdn
> = 'cdn1' AND time > '2017-05-03T09:22:00-06:00' AND time <
> '2017-05-04T09:22:00-06:00' GROUP BY time(60s), cdn ORDER BY asc
> summary query:
> SELECT mean(value), percentile(value, 5), percentile(value, 95),
> percentile(value, 98), min(value), max(value), sum(value), count(value) FROM
> "monthly"."metricName.cdn.1min" WHERE cdn = 'cdn1' AND time >
> '2017-05-03T09:22:00-06:00' AND time < '2017-05-04T09:22:00-06:00' GROUP BY
> time(60s), cdn
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
