Closed
Bug 1124371
Opened 10 years ago
Closed 9 years ago
Loop Reporting JSON that includes rooms
Categories
(Cloud Services :: Operations: Metrics/Monitoring, task)
Tracking
(Not tracked)
RESOLVED
WONTFIX
People
(Reporter: kparlante, Assigned: whd)
References
Details
We need to revist the "callers" metrics, given that there's a new rooms api, so a new way to be a successful caller. As before, the loop server doesn't observe the success of the media, so we're only measuring success in connecting the two users.
Some definitions for our approach:
- Unique users ==> unique hawk ids (uid, calleeId)
- Successful call user ==> same as our "caller" metrics before. This counts both direct and call-url callers.
- Successful room user ==> users who have been in a conversation/room with another user.
- Successful session ==> successful calls (same as before) + successful room sessions. For rooms, this means the number of times a person has joined a room with another person already in it. (This will be an overcount; we have unique rooms as part of the rooms flow).
Reporter | ||
Comment 1•10 years ago
|
||
To start, lets do parallel "rooms" counts for all of the caller metrics (we can skip fortnightly):
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_new_callers.json
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_cumulative_callers.json
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_daily_unique_callers.json
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_weekly_recurring.json
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_monthly_recurring.json
In each case, we care about unique uid's for this query:
method:post AND path:rooms AND action:join AND errno:0 AND participants:2
Also, a "room sessions", which is the moral equivalent of "call setups" (only we don't have parallel failure data). The proposed approach is to just count:
method:post AND path:rooms AND action:join AND errno:0 AND participants:2
Assignee | ||
Comment 2•10 years ago
|
||
I'm going to append "rooms" to the names for the rooms counts e.g.
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_monthly_recurring_rooms.json
Do we want cumulative callers etc. to start from 0 or from the current value? I'm guessing 0.
I'm going to put the rooms sessions data at:
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_rooms_sessions.json
Assignee | ||
Comment 3•10 years ago
|
||
The following filters have been set up, and should start to see activity once 0.15.1 goes to prod (possibly tomorrow):
https://heka.shared.us-west-2.prod.mozaws.net/#plugins/filters/LoopRoomsSessionsMetrics (json)
https://heka.shared.us-west-2.prod.mozaws.net/#plugins/filters/LoopRoomsSessionsCount (cbuf of same)
https://heka.shared.us-west-2.prod.mozaws.net/#plugins/filters/LoopTotalUsersWithRooms
https://heka.shared.us-west-2.prod.mozaws.net/#plugins/filters/LoopActiveDailyCallersWithRooms
Recurring counts is an offline computation but the code is the same (only the message matcher differs). Once we have some data those numbers should be available at:
https://heka.shared.us-west-2.prod.mozaws.net/#plugins/filters/LoopRoomsMetrics
Reporter | ||
Comment 4•10 years ago
|
||
Looks like 0.15.1 landed in production (I'm seeing "participants" in kibana), so should be able to add this to the dashboard tomorrow.
Reporter | ||
Comment 5•10 years ago
|
||
(In reply to Katie Parlante from comment #1)
> In each case, we care about unique uid's for this query:
> method:post AND path:rooms AND action:join AND errno:0 AND participants:2
>
> Also, a "room sessions", which is the moral equivalent of "call setups"
> (only we don't have parallel failure data). The proposed approach is to just
> count:
> method:post AND path:rooms AND action:join AND errno:0 AND participants:2
Looking at the actual data in kibana, looks like "participants" is the number of people in the room when you try to join. So these queries should change to:
method:post AND path:rooms AND action:join AND errno:0 AND participants:1
Comment 6•10 years ago
|
||
Yes true, this might be something we want to change, should we?
https://github.com/mozilla-services/loop-server/blob/master/loop/middlewares.js#L90-L92
Reporter | ||
Comment 7•10 years ago
|
||
(In reply to Rémy Hubscher (:natim) from comment #6)
> Yes true, this might be something we want to change, should we?
>
> https://github.com/mozilla-services/loop-server/blob/master/loop/middlewares.
> js#L90-L92
No need to change.
Reporter | ||
Comment 8•10 years ago
|
||
Ok, I have more use cases working locally with the production data, but I'm not pushing live tonight -- the combination of so few data points & the load testing snafu renders it more confusing than useful.
Changes still needed:
- Add the "Fields[user_agent_browser] != NIL" filter to the rooms variants & reprocess
- Make 7 day & 30 day active (not recurring) available to the dashboard (for both calls and rooms)
- Make room session json available to the dashboard
Assignee | ||
Comment 9•10 years ago
|
||
"Fields[user_agent_browser] != NIL" has been added to the live filters, but still needs reprocessing.
Today I wrote a generic "actives" filter at https://github.com/mozilla-services/puppet-config/blob/loop_rooms_metrics/shared/modules/shared/files/hekad/lua_filters/active_counts.lua which I'll configure for 7 & 30 day actives. I'll also do the reprocessing and dashboard proxy rules tomorrow.
Assignee | ||
Comment 10•10 years ago
|
||
Rooms variants should be recomputed and running live. The 7 day and 30 day counts will not show up until 7 days and 30 days have passed, but the other data should look better. When the 7 and 30 day counts are available, they will be at:
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_room_weekly_unique_callers.json
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_room_monthly_unique_callers.json
I also tweaked the dashboard name for the rooms variants for consistency (s/rooms/room/); where they were:
/loop-server-dashboard/data/loop_rooms_$METRIC.json
They are now:
/loop-server-dashboard/data/loop_room_$METRIC.json
7 and 30 day counts for "calls" are also live and backfilled with data from 2014-12; that data is available at:
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_weekly_unique_callers.json
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_monthly_unique_callers.json
Session JSON available at:
https://metrics.services.mozilla.com/loop-server-dashboard/data/loop_room_sessions.json
(feel free to suggest better names for any of the above)
Per discussion in bug #1122506 I'm going to modify "Fields[user_agent_browser] != NIL" to be "Fields[agent] != $LOADS_USER_AGENT" for all cases where we use the former, which I will do tomorrow.
Assignee | ||
Updated•9 years ago
|
Status: NEW → RESOLVED
Closed: 9 years ago
Resolution: --- → WONTFIX
You need to log in
before you can comment on or make changes to this bug.
Description
•