Glean crash pings are at parity with those from Telemetry

[Alex Franchuk]

Now that Bug 1815316 - Add additional metrics to Glean crash ping to reach parity with Telemetry has landed, the crash pings sent by Glean have all the relevant information that Telemetry crash pings provide.

This is for Desktop only at the moment, however we will soon fill out Android crash pings too (which currently are a blind spot for us).

To begin deprecating the Telemetry crash ping, I'd like to be aware of all post-processing that is done on the data. It should be somewhat mechanical to convert to using the Glean data instead.

Please let me know if you use or post-process the Telemetry crash ping data. I already know that https://github.com/mozilla/process-top-crashes uses it, and I expect there are stability dashboards in Looker/STMO as well. Thanks!

[Tom Ritter]

Just to make sure I understand correctly - these crash pings are still
crash pings that crash, right? I know there's been a lot of talk
about 'crash pings that don't actually crash' and I'm hopeful we'll
get those some day, but this isn't that, right?

-tom

> --
> You received this message because you are subscribed to the Google Groups "crash-reporting-wg" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to crash-reporting...@mozilla.com.
> To view this discussion on the web visit https://groups.google.com/a/mozilla.com/d/msgid/crash-reporting-wg/1301f6b5-8a12-4d04-80de-95e74cdc24a6n%40mozilla.com.
> For more options, visit https://groups.google.com/a/mozilla.com/d/optout.

[Gabriele Svelto]

On 06/06/24 16:46, Tom Ritter wrote:
> Just to make sure I understand correctly - these crash pings are still
> crash pings that crash, right? I know there's been a lot of talk
> about 'crash pings that don't actually crash' and I'm hopeful we'll
> get those some day, but this isn't that, right?

Yes, but we've deliberately designed the types they hold so that we'll
eventually be able to send stacks, whether they represent crashes or
not. This is a big step towards having that (and of course, allows us to
get rid of yet another legacy ping).

Gabriele

[Alex Franchuk]

Yes, these are designed so that we can manually send stacks at interesting/error points without crashing in the future. Also note that we still need to support Glean crash pings in the crash reporter (we will likely add it to the pingsender program), see Bug 1821091 - Send Glean crash pings in the crash reporter, so main process crashes may be under-reported right now (but bug 1821091 is a high priority because of that).

-Alex

[William Kahn-Greene]

I put together this brief on crash pings symbolication and signature generation. It includes prior art which includes systems that are using the current legacy crash pings.

https://docs.google.com/document/d/1Won8mmsq8HRhSYWk9fCTm7geeEZxrA35PjnG-COKnbQ/edit#

I don't think that covers everything, but it probably gives you some threads to follow up on.

/will

--
You received this message because you are subscribed to the Google Groups "crash-reporting-wg" group.
To unsubscribe from this group and stop receiving emails from it, send an email to crash-reporting...@mozilla.com.

To view this discussion on the web visit https://groups.google.com/a/mozilla.com/d/msgid/crash-reporting-wg/c2749352-9918-491f-973e-868ba00e4dd4%40mozilla.com.