Aggregate stats from a project

[woodychuck]

I'd like to use the client to get aggregate stats for projects from our OpenShift platform. However query_metrics_stats() insists on a metric_id while I rather just want to filter by tags to get my result. I reckon a simple if statement enabling to use metric_id=None could resolve this but of course being new to this stuff leaves some uncertainty. Anyway, the following change in metrics.py did the trick for me:

    def _get_single_id_url(self, previous_url, id):
        if id:
            return previous_url + '/{0}'.format(HawkularBaseClient.quote(id))
        else:
            return previous_url

Does this make sense or am I supposed to use the client in a different way to retrieve my stats?

[burmanm]

Do you want them as separate ones on your end (for further processing) or merged together at Hawkular-Metrics' end?

[woodychuck]

In the end I just want aggregate (stacked) stats for complete projects in a TBD number of buckets, e.g. GET /hawkular/metrics/gauges/stats?tags=descriptor_name%3Acpu%2Frequest%2Ctype%3Apod&stacked=true&buckets=30&start=$(date -d -30minutes +%s%3N)" (in order to compare to cpu/usage_rate, get the max and - along with memory usage - calculate the billing parameters). If I wanted separate metrics for each pod I figured I need to get the metrics list and iterate through it (which can be cumbersome given cron job pods cumulating up etc.)

[burmanm]

I'll create a version that allows this (given enough new Openshift-metrics)

[burmanm]

There's a version now available in PyPi (0.5.4) that allows leaving metric_id as None

[woodychuck]

Hi, thanks that's great! However it seems to be missing hawkular/alerts/ in the package?

[burmanm]

So it would appear, I'll create a version with the old alerts-package also. I'll track this in #57

[burmanm]

0.5.5 is now in the PyPi

[woodychuck]

Lovely, works like charm, thanks again!