Closed Bug 1336376 Opened 8 years ago Closed 8 years ago

Expand the range of values for URLCLASSIFIER_LOOKUP_TIME

Categories

(Toolkit :: Safe Browsing, defect, P2)

defect

Tracking

()

RESOLVED FIXED
mozilla54
Tracking Status
firefox54 --- fixed

People

(Reporter: francois, Assigned: francois)

Details

Attachments

(1 file)

Right now, URLCLASSIFIER_LOOKUP_TIME is defined like this: "kind": "exponential", "high": 500, "n_buckets": 10, which unfortunately hides how bad lookup time can be by putting everything over 500 ms in the same bucket. We should bump the number of buckets and move the high to at least 5 seconds.
Benjamin, is there a problem with increasing the number of buckets and/or the high value of an existing probe? https://developer.mozilla.org/en-US/docs/Mozilla/Performance/Adding_a_new_Telemetry_probe points out that n_values can't be increased later for technical reasons, but is silent when it comes to n_buckets. The only thing I could find in there is a recommendation that a new probe be created instead, but no explanation is given.
Flags: needinfo?(benjamin)
Aggregation and t.m.o will break if you change a histogram like this without changing the name. Since you're changing the bucket boundaries anyway, the best practice is to change the name to URLCLASSIFIER_LOOKUP_TIME_2 and so forth.
Flags: needinfo?(benjamin)
Comment on attachment 8833273 [details] Bug 1336376 - Expand the range of values for URLCLASSIFIER_LOOKUP_TIME. https://reviewboard.mozilla.org/r/109526/#review110574
Attachment #8833273 - Flags: review?(benjamin)
(In reply to Benjamin Smedberg [:bsmedberg] from comment #3) > Aggregation and t.m.o will break if you change a histogram like this without > changing the name. Since you're changing the bucket boundaries anyway, the > best practice is to change the name to URLCLASSIFIER_LOOKUP_TIME_2 and so > forth. Can I delete the original probe (i.e. replace it with LOOKUP_TIME_2) or is there a technical reason why I should leave it around for a while?
Flags: needinfo?(benjamin)
I *think* we stop aggregating and displaying the old metric on t.m.o as soon as it's removed from nightly, but I'm not 100% sure of that. Ryan do you know?
Flags: needinfo?(benjamin) → needinfo?(rharter)
Comment on attachment 8833273 [details] Bug 1336376 - Expand the range of values for URLCLASSIFIER_LOOKUP_TIME. https://reviewboard.mozilla.org/r/109526/#review111072
Attachment #8833273 - Flags: review?(benjamin) → review+
I'll let Frank answer this; He has more experience with t.m.o.
Flags: needinfo?(rharter) → needinfo?(fbertsch)
TLDR; You can go ahead and replace it, won't hurt a thing. Longer story: On the backend, we aggregate anything we see, regardless of whether we expected it for a certain channel/version/build. Only on the frontend (with the aggregates service) do we try to look up what the definition is for a probe. If this lookup fails, you'll get a 404. Basically, if the probe is no longer available in histograms.json or scalars.yaml, it's not available on TMO. This has it's drawbacks (historical probes aren't available), but hasn't failed us so far.
Flags: needinfo?(fbertsch)
Pushed by fmarier@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/3a6b42b3d233 Expand the range of values for URLCLASSIFIER_LOOKUP_TIME. r=bsmedberg
I had to back this out in https://hg.mozilla.org/integration/autoland/rev/2ff48b77d380c9c649d2f5f1d54b4f5c5b3340ad for build failures like https://treeherder.mozilla.org/logviewer.html#?job_id=74977218&repo=autoland It looks like an easy fix, but I don't know enough about the whitelist to know what removing it would mean elsewhere.
Flags: needinfo?(francois)
Pushed by fmarier@mozilla.com: https://hg.mozilla.org/integration/autoland/rev/017a3a0d4a00 Expand the range of values for URLCLASSIFIER_LOOKUP_TIME. r=bsmedberg
Flags: needinfo?(francois)
Status: ASSIGNED → RESOLVED
Closed: 8 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla54
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: