1034095 - Decrease in Telemetry submission rate on aurora channel

Status: RESOLVED FIXED

(Toolkit :: Telemetry, defect)

Description

[Mark Reid [:mreid]]

Attachment: aurora submission rate.png

The submission rate on the aurora channel has dropped noticeably since about June 28th.

Like bug 1030217, it may be due to bug 1007723, but the decrease is more pronounced (see attachment).

Are there any known reasons why the rate would have decreased at this time?

Comment 1

[Mark Reid [:mreid]]

Attachment: aurora submission rate.png

Previous attachment showed the data volume rather than the submission rate itself.  The shape is the same, but the scale is different.

Comment 2

[Vladan Djeric (:vladan)]

Looking at the commit log for the toolkit/components/telemetry directory on Aurora:

a) On June 26, bug 1017806: Telemetry Experiments: builtin support for multiple branches of an experiment
b) On June, bug 1007710: Add information about the currently-active experiment to the telemetry ping, r=gfritzsche+vladan

After glancing at the patches, neither seems like a likely candidate.

Comment 3

[Vladan Djeric (:vladan)]

Could this be bug 1030161? The A-Team stopped Talos submitting Telemetry starting June 24th

Comment 4

[Mark Reid [:mreid]]

Attachment: aurora-submission-rate-2014-07-29.png

Looking at the graph for Aurora TELEMETRY_FILES_EVICTED today, we see the following:
http://telemetry.mozilla.org/#filter=aurora%2F33%2FTELEMETRY_FILES_EVICTED

 # discarded
  per ping     submissions     # discarded 
 -----------   -----------    -----------
    0           37050          0   
    1           4020           4020    
    2           677            1354    
    3           190            570 
    4           100            400 
    5           55             275 
    6           30             180 
    7           28             196 
    8           26             208 
    9           21             189 
    10          16             160 
    11          12             132 
    12          19             228 
    13          17             221 
    14          11             154 
    15          9              135 
    16          15             240 
    17          2              34  
    18          8              144 
    19          8              152 
    20          3              60  
    21          5              105 
    22          8              176 
    23          8              184 
    24          9              216 
    25          7              175 
    26          9              234 
    27          2              54  
    28          1              28  
    29          1              29  
    30+         59             1770 (at least)
-----------------------------------------------
                42426          12023

So the total number of expected pings without file eviction is 42426 + 12023 = 54449, and the fraction that were discarded is 12023 / 54449 = 0.2208

A 22% decrease appears to me to match with the observed decrease in the submission rate (see updated graph attached).

Comment 5

[Mark Reid [:mreid]]

I think we can safely say that the observed decrease is due to the same reason as bug 1030217.