1388310 - Validate incoming health ping data (nightly)

Status: RESOLVED FIXED

(Toolkit :: Telemetry, defect, P1)

Description

[Kate Ustiuzhanina]

This include:
1. Validate input data.
2. Compute Health pings per client.
3. Compute Health pings per day. 
4. Compute Health pings per client per day. 
5. Show sendFailure and pingDiscardedForSize distribution.

Comment 1

[Kate Ustiuzhanina]

Georg, could you, please, make a review?
PR https://github.com/mozilla/mozilla-reports/pull/72

Comment 2

[Georg Fritzsche [:gfritzsche]]

(In reply to Kate Ustiuzhanina from comment #1)
> Georg, could you, please, make a review?
> PR https://github.com/mozilla/mozilla-reports/pull/72

Chris, a heads-up on this report. This is relevant to what you are looking into as well.

Comment 3

[Chris H-C :chutten]

I look forward to seeing an updated report with your changes included.

I would appreciate a note about possible reasons the data from the health ping might differ from data included in the main ping. For instance, the difference in proportions of failure types (https://mzl.la/2uUVUwy) which might just be a matter of an even more pronounced time-of-arrival effect as per bug 1371312 comment 0

Comment 4

[Kate Ustiuzhanina]

Hi Chris. 
About differences in numbers. One big reason is that we have published health pung in nightly 56 around 26 June, so after filtering the numbers looks similar, but still not the same. (https://telemetry.mozilla.org/new-pipeline/dist.html#!cumulative=0&end_date=2017-08-01&keys=__none__!__none__!__none__&max_channel_version=nightly%252F56&measure=TELEMETRY_SEND_FAILURE_TYPE&min_channel_version=null&processType=*&product=Firefox&sanitize=1&sort_keys=submissions&start_date=2017-07-26&table=1&trim=1&use_submission_date=0)

Comment 5

[Kate Ustiuzhanina]

Hi Chris, I updated PR (https://github.com/mozilla/mozilla-reports/pull/72), could you, please, take a look?

Comment 6

[Kate Ustiuzhanina]

(In reply to Kate Ustiuzhanina from comment #5)
> Hi Chris, I updated PR (https://github.com/mozilla/mozilla-reports/pull/72),
> could you, please, take a look?

I updated code according your comments (or made some answers).

Comment 7

[Kate Ustiuzhanina]

Fixed in PR https://github.com/mozilla/mozilla-reports/pull/72