We need a new metric type for coarse time measurements
Categories
(Data Platform and Tools :: Glean Metric Types, enhancement, P3)
Tracking
(Not tracked)
People
(Reporter: Dexter, Assigned: mdroettboom)
References
Details
(Whiteboard: [telemetry:glean-rs:backlog][new-metric])
Attachments
(1 file)
Proposal for changing an existing or adding a new Glean metric type
Who is the individual/team requesting this change?
Daosheng Mu, Firefox Reality
Is this about changing an existing metric type or creating a new one?
Creating a new metric type.
Can you describe the data that needs to be recorded?
We need to measure coarser times, such as "how long was the tab open?": this could span from a few minutes to days (for "user-time" centered measurements).
Can you provide a raw sample of the data that needs to be recorded (this is in the abstract, and not any particular implementation details about its representation in the payload or the database)
It's a distribution of samples, e.g. a tab was opened for 5 hours, another for 1 hour, an additional one for 4 days. What really matters is not the specific value, but the shape of the distribution.
What is the business question/use-case that requires the data to be recorded?
We need to understand what is the distribution of tab life across our users.
How would the data be consumed?
We would like to see aggregated counts available on the GLAM dashboard or, in the meanwhile, through custom SQL queries.
Why existing metric types are not enough?
The Glean SDK supports the Timing Distribution metric type, which allows to measure times under 10 minutes with great nanoseconds-precision (for "cpu-time" centered measurements) . The lifetime of a browser tab can be much longer than that.
What is the timeline by which the data needs to be collected?
TBD.
Reporter | ||
Comment 1•5 years ago
|
||
Note: this came up during the data migration planning for Firefox Reality.
Comment 2•5 years ago
|
||
I am curious if we can just use TimeSpan
that we don't need to care about the limitation of 10 mins from Time Distribution
.
Comment 3•5 years ago
|
||
If we're only recording a single span of time we should of course use a Timespan. But, taking Alessio's example of how long tabs are open... for that we need multiple timespans and are interested in their distribution. That's normally what a time distribution would do, if it weren't for that 10min limit.
Reporter | ||
Updated•5 years ago
|
Reporter | ||
Comment 4•5 years ago
|
||
I reframed comment 0 to conform to the new template, and moved this bug to the right component.
Daosheng, are the information in comment 0 correct? Is this still needed? This can now move as we finalized the process for adding new metric types.
Comment 5•5 years ago
|
||
(In reply to Alessio Placitelli [:Dexter] from comment #4)
I reframed comment 0 to conform to the new template, and moved this bug to the right component.
Daosheng, are the information in comment 0 correct? Is this still needed? This can now move as we finalized the process for adding new metric types.
It looks great to me. This is what we're looking for.
Reporter | ||
Comment 6•5 years ago
|
||
Here's the document for discussing the addition of this new metric type related to this bug, as highlighted in the process document.
I'm flagging all of you for review on the request. Please examine it and see if there's any blocker from your point of view.
Feedback is due by March 2nd, 2020.
Note This is the first run of the process, so please be patient and do not hesitate to leave feedback about the process too!
Updated•5 years ago
|
Assignee | ||
Comment 7•5 years ago
|
||
I added the start of a concrete proposal to the doc, since it was kind of missing (we're learning about the holes is this process as we go). We can probably have the back and forth in the doc, and once more final, will give the rest of use something concrete to consider.
Comment 8•5 years ago
|
||
I left a comment on the doc.
One minor nit regarding the document: it would be nice, when possible, to not copy paste the markdown to gdocs as it doesn't format that properly. I find it hard to scan through the document without the proper formatting :) That's easily fixable by copying the parsed content from the bugzilla comment.
Comment 9•5 years ago
|
||
Comment 10•5 years ago
•
|
||
Comment 11•5 years ago
|
||
Assignee | ||
Comment 12•5 years ago
|
||
Thanks Marina. I'll address your comment about inconsistency in the doc.
Comment 13•5 years ago
|
||
Updated•5 years ago
|
Assignee | ||
Updated•5 years ago
|
Assignee | ||
Comment 14•5 years ago
|
||
Approved the proposal, and created bug 1630997 for the implementation.
Assignee | ||
Updated•5 years ago
|
Updated•4 years ago
|
Reporter | ||
Updated•3 years ago
|
Reporter | ||
Updated•3 years ago
|
Description
•