Closed Bug 808003 Opened 7 years ago Closed 7 years ago

Add telemetry probes for tab zombification

Categories

(Firefox for Android :: General, defect)

All
Android
defect
Not set

Tracking

()

RESOLVED FIXED
Firefox 19

People

(Reporter: kats, Assigned: kats)

References

Details

Attachments

(2 files)

We should add some telemetry probes for tab zombification, both via memory-pressure (as added in bug 784040) and tab expiration (as added in bug 792143/bug 803575).
Adding some telemetry data for tab zombification. I believe these data points should help us tune the tab expiry behaviour and also give us a little bit more information about usage patterns in the wild.
Attachment #678402 - Flags: review?(mark.finkle)
Attachment #678402 - Flags: review?(gpascutto)
Attachment #678399 - Flags: review?(gpascutto) → review+
Comment on attachment 678402 [details] [diff] [review]
(2/2) Add telemetry histograms for tab zombification

Looks good to me
Attachment #678402 - Flags: review?(mark.finkle) → review+
Comment on attachment 678402 [details] [diff] [review]
(2/2) Add telemetry histograms for tab zombification

Review of attachment 678402 [details] [diff] [review]:
-----------------------------------------------------------------

LGTB but see remarks, might want to shuffle your buckets.

::: toolkit/components/telemetry/Histograms.json
@@ +2281,5 @@
>    },
> +  "FENNEC_LOWMEM_TAB_COUNT": {
> +    "kind": "linear",
> +    "high": 100,
> +    "n_buckets": 50,

I think you're probably much more interested if there are 1,2,3,4... tabs when we go into lowmem rather than if there are 70 or 71 tabs, so I'd change this to an exponential, which would also allow you to lower the buckets to 20 or 30ish and still get better coverage in the range you're interested in.

@@ +2323,5 @@
>    },
> +  "FENNEC_TAB_EXPIRED": {
> +    "kind": "exponential",
> +    "low": 10,
> +    "high": 1000000,

Do you really need a range of 12 days? I know the range is exponential but trimming it will still give you more buckets in the interesting range as above.
Attachment #678402 - Flags: review?(gpascutto) → review+
(In reply to Gian-Carlo Pascutto (:gcp) from comment #4)
> I think you're probably much more interested if there are 1,2,3,4... tabs
> when we go into lowmem rather than if there are 70 or 71 tabs, so I'd change
> this to an exponential, which would also allow you to lower the buckets to
> 20 or 30ish and still get better coverage in the range you're interested in.

True, I changed this to be exponential with 30 buckets.

> Do you really need a range of 12 days? I know the range is exponential but
> trimming it will still give you more buckets in the interesting range as
> above.

Dropped the range to 7 days after verifying that anything beyond the range will just get dumped into the last bucket.

https://hg.mozilla.org/integration/mozilla-inbound/rev/03393060a070
https://hg.mozilla.org/integration/mozilla-inbound/rev/4f5f601bc741
Also https://hg.mozilla.org/integration/mozilla-inbound/rev/e594a14532be

(gcp pointed out that I only updated one of the two 12-day histograms, so this updates the other one. r=gcp via IRC)
You need to log in before you can comment on or make changes to this bug.