Add groupby aggregation function

[Sergio Colinas]

When you aggregate measures using groupby, you can't specify an aggregation function for the grouping aggregation (it will use the same as the measures).

For example, if you want to get the average memory usage groupby project in a month, you'll need the sum of each instance average memory usage.

If you use sum you will get the sum of each instance memory usage for each min granularity in a month, this will duplicate the measures and the result will be wrong. If you use mean, you will get the average instance memory usage, not the project average.

[Julien Danjou (jdanjou)]

I'm not sure I understand. If you want to the average memory usage of instances of a project, you just aggregate the 'mean' of all instances grouped by projects.

So asking the aggregation of 'mean' for instances/memory grouped by project should give you a list of the average of each project' instances memory usage over different period of time. You then just need to re-average that yourself is needed.

What do I miss?

[Sergio Colinas (sergio-n)]

Hi Julien,

I think there are two different aggregation functions. On the one hand you have the aggregation function of the data itself (specified in the archive policy), and in the other hand the one used for the group by. You should be able to combine them in the way you need.

For example, if you have two instances in a project, one using 1gb average and the other 2gb. If you ask for 'mean' you will get 1.5gb.

Instead of that I would like to get the sum of the average which is 3gb but I can only specify one aggregation method.

[Sergio Colinas (sergio-n)]

Hi Julien,

Did you understand the issue? I think it should be pretty easy to add this feature in gnocchi and it will give us the ability to query the measures using independent aggregation functions for group by and raw data aggregation.

[Julien Danjou (jdanjou)]

Understood, changed the status to triaged! Thanks Sergio!

[OpenStack Infra (hudson-openstack)]

Related fix proposed to branch: master
Review: https://review.openstack.org/350043

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/350043
Committed: https://git.openstack.org/cgit/openstack/gnocchi/commit/?id=c13a21a8ae693adb4ff2f25bbf4d1908c0eb84dc
Submitter: Jenkins
Branch: master

commit c13a21a8ae693adb4ff2f25bbf4d1908c0eb84dc
Author: Julien Danjou <email address hidden>
Date: Tue Aug 2 15:16:56 2016 +0200

    storage: allow to specify regaggregation over aggregation retrieval

    Currently, when retrieving e.g. 'mean' aggregation of several metrics,
    the aggregation of those metrics is always done using 'mean'.
    This patches allows to specify a reaggregation methods, which applies on
    top of the measures retrieved.

    You can know retrieve e.g. the 'max' of the 'mean' of several metrics.

    Related-Bug: #1573023
    Change-Id: I9ea60de498e6f81c738a7236a392019b47b0c269

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/352023

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/352023
Committed: https://git.openstack.org/cgit/openstack/gnocchi/commit/?id=82509d782e8a21d70a848065f4948f98419aca54
Submitter: Jenkins
Branch: master

commit 82509d782e8a21d70a848065f4948f98419aca54
Author: Julien Danjou <email address hidden>
Date: Sat Aug 6 15:48:17 2016 +0200

    rest: allow to specify reaggregation methods in aggregation

    Change-Id: Ib37c6b7e8b8214bd8b3e8f2e92547fed56923374
    Closes-Bug: #1573023