1207557 - Fix the Remote Debugging Protocol telemetry probes.

Status: RESOLVED WONTFIX

(DevTools :: Debugger, defect)

Description

[Jan Keromnes [:janx]]

As Allison mentioned in bug 1196785 comment 39, we should:
- Use an explicit expiration version (not "never")
- Set n_buckets to 100 or lower.

Currently, all our "DEVTOOLS_DEBUGGER_RDP_.*" probes in [1] use expiration "never" and 1000 buckets. We should fix that.

[1] https://dxr.mozilla.org/mozilla-central/source/toolkit/components/telemetry/Histograms.json#5649

Comment 1

[Nick Fitzgerald [:fitzgen] [⏰PST; UTC-8]]

:bsmedberg said number of buckets doesn't have too much of an affect on packet size, so why does it matter?

https://bugzilla.mozilla.org/show_bug.cgi?id=1203590#c4

Comment 2

[Allison Naaktgeboren :ally]

(In reply to Nick Fitzgerald [:fitzgen][:nf] from comment #1)
> :bsmedberg said number of buckets doesn't have too much of an affect on
> packet size, so why does it matter?
> 
> https://bugzilla.mozilla.org/show_bug.cgi?id=1203590#c4

:vladan mentioned it to me. Iirc, there's impact server-side.

Comment 3

[Vladan Djeric (:vladan)]

We use a sparse histogram bucket representation when sending the Telemetry payload, but the histogram code we inherited from Chromium instantiates all histogram buckets in memory, so the bucket count ends up affecting runtime memory use (important on Fennec).

On the server-side, histograms with thousands of buckets increase storage costs, aggregation times, etc.

Comment 4

[Vladan Djeric (:vladan)]

Btw, you'll have to create new histograms (with new names) if you want to change any of the histogram parameters that affecting bucketing

Comment 5

[Jan Keromnes [:janx]]

Fixing up the histogram parameters is probably not worth the cost of creating all new histograms. Closing as "won't fix".

Benjamin, Allison and Vladan's replies on n_buckets seem to contradict your earlier statement referred to in comment 1. Needinfo-ing you just so that you're aware this discussion happened.