mozilla / glam

Mozilla's primary interactive dashboard for examining the distribution of telemetry values.
https://glam.telemetry.mozilla.org
Mozilla Public License 2.0
20 stars 23 forks source link

GLAM breaks when Normalization none is selected for a probe that doesn't have non-normalized data #2777

Closed edugfilho closed 3 months ago

edugfilho commented 6 months ago

Spin off of https://github.com/mozilla/glam/issues/2774 STR:

Navigate to https://glam.telemetry.mozilla.org/
select telemetry probe 'gc_minor_us'
select channel 'release'
select normalization 'none'

Result:

The site becomes unresponsive. The console reports the following exception was thrown:


    l QuantileExplorerView.svelte:77
    update QuantileExplorerView.svelte:89
    xt index.mjs:1343
    _t index.mjs:1307```
edugfilho commented 6 months ago

This probe doesn't have enough client count on release to provide for accurate statistics. GLAM was never supposed to have collected release data for the probe in question, due to the low client counts. The reason why there are two builds there is because those builds were aggregated before a better client count filtering was put in place for release.

Considering that, I removed the few data points for that probe in release from the dataset, which is now making GLAM display a "Not Found" message when the probe is selected for release instead of breaking.

jonco3 commented 6 months ago

I checked and this now works for the gc_minor_us probe. But this still affects other probes.

For example, if I look at gc_mark_ms and select channel 'release' and then then normalization 'none', the following exception is thrown:

Uncaught (in promise) TypeError: n[(n.length - 1)][Gn(...)].map is not a function
    l https://glam.telemetry.mozilla.org/static/bundle.js:9
    update https://glam.telemetry.mozilla.org/static/bundle.js:9
    xt https://glam.telemetry.mozilla.org/static/bundle.js:1
    _t https://glam.telemetry.mozilla.org/static/bundle.js:1

Would it be possible to check for this in the front end?

edugfilho commented 6 months ago

Thanks @jonco3 for showing that wasn't an isolated issue. I'm re-opening the ticket.

Iinh commented 4 months ago

I can take this work @edugfilho