Closed Bug 947294 Opened 12 years ago Closed 8 years ago

Use update-timer instead of idle-daily to schedule Telemetry pings

Categories

(Toolkit :: Telemetry, defect)

x86
All
defect
Not set
normal

Tracking

()

RESOLVED FIXED

People

(Reporter: mconley, Assigned: mconley)

References

Details

(Whiteboard: [Australis:P-])

Split off from bug 863872. The idle-daily notification is sketchy, and its occurrence is too rare. We want to use a different mechanism in the hopes of increasing our Telemetry ping intake volume. The Blocklist updater is triggered via nsIUpdateTimerManager over a 24hr period. I think we can do the same.
Since we're not depending on "idle" anymore, we'll need to be careful not to create races between update-timer send, async saved-session ping loading, and forced overdue ping sends on startup
We also want to be sure that gather-telemetry (currently sent on idle-daily) doesn't cause unwanted jank / battery drainage.
I think idle-daily timer stays scheduled even when there are no listeners, but that would be worth looking in to. On mobile we listen for the application to go into the background and save a session (in case the OS kills us to free up memory). We could cancel our timers and listen for a foreground notification to start them up again, but I think this is going to be a tiny effect compared to all the other background events the browser runs. We could add telemetry around how long it takes to send gather-telemetry, though that only catches observers that do their gathering synchronously.
Whiteboard: [Australis:P-]
Hey Irving / Vladan, How concerned are we with this bug still? Looking at the incoming Telemetry Ping logs, I don't think the overdue flush really did much to increase our incoming pings... Are we still in a state where the number of pings that are coming in are not enough to paint an accurate enough picture (as accurate a picture that Telemetry can deliver, anyhow)?
Flags: needinfo?(vdjeric)
Flags: needinfo?(irving)
Good heavens, http://telemetry.mozilla.org/#nightly/29/SIMPLE_MEASURES_PINGSOVERDUE (in Table view) shows one client that had somewhere between 462 and 564 overdue pings, and http://telemetry.mozilla.org/#nightly/29/SIMPLE_MEASURES_SAVEDPINGS shows 4 with > 1.5k saved pings... Looking at subsets of the PINGSOVERDUE and PINGSDISCARDED data by date range, it looks like we are improving in how well we push data to the server - the % of clients that have old pings is going down. I'd still like to have this change made, but it's not a really high priority for me.
Flags: needinfo?(irving)
idle-daily is not currently in a good shape on server-side and it's not being used for anything, so let's shelve this for now
Flags: needinfo?(vdjeric)
With Unification, this is no longer relevant.
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.