Ensure that the sampling for this instrument is per session not per-pageview.

mediawiki.web_ui_actions uses

'sample' => [
  'unit' => 'session',
  'rate' => 0.2,
],

Source: https://noc.wikimedia.org/conf/highlight.php?file=ext-EventStreamConfig.php

Yanking the sampling logic out of the instrument should suffice (in theory).

@mpopov Does removing the sampling logic leave the configuration without a default? Or is there another place where the default should be set?

@KSarabia-WMF: The default is in the stream config (pasted above) and it's currently set to 20% of sessions on all projects. And what's happening with the migrated instrumentation is that it's being double-sampled – the instrument initializes on 20% of pageviews (1% on English projects) and when it produces events to the mediawiki.web_ui_actions stream, only 20% of sessions will actually submit an event.

According to https://wikitech.wikimedia.org/wiki/Metrics_Platform/How_to/Creating_a_Stream_Configuration#Example_1 you would probably want to add something like

'+enwiki' => [
        'mediawiki.web_ui_actions' => [
            'sample' => [
                'rate' => 0.01,
                'unit' => 'session',
            ],
        ],
    ],

so that we're only collecting data from 1% of sessions on English Wikipedia. Volume of data is not a problem right now because of the double-sampling but we would want to add this.

@cjming: Can you please confirm? (That this is how we would set an enwiki-specific sampling rate.) Also! would '+en' work? The docs mention specifying deployment groups but I don't know if language groups are supported.

Confirming that overrides per wiki should be added to config the way @mpopov has described.

Fwiw, there's already an override for enwiki and mediawiki.web_ui_actions in prod config.

I'm a little confused tho -- there's mediawiki.web_ui_actions which is not using MP schemas and there's mediawiki.web_ui_scroll_migrated which is using MP's web base schema. If the data QA parity checks between the two have been done and verified, one of these should probably go away.

@KSarabia-WMF overriding either is the same mechanism as Mikhail noted above but maybe mediawiki.web_ui_actions should be retired in favor of mediawiki.web_ui_actions_migrated if the goal is to have a single table for UI actions that uses MP?

One thing I might suggest after all this is work is wrapped up is to clean up the corresponding Hive tables after the older stream configs have been removed as a follow up ticket (they don't get automatically deleted when stream configs are removed afaik).

would '+en' work? The docs mention specifying deployment groups but I don't know if language groups are supported.

@mpopov '+en' will not work - dblist is hardcoded - all language codes have wiki appended and deployment groups are for train rollouts i.e. group2 which is where enwiki is included.

I believe mediawiki.web_ui_actions succeeds eventlogging_DesktopWebUIActionsTrackingand eventlogging_MobileWebUIActionsTracking
and mediawiki.web_ui_scroll_migrated succeeds mediawiki.web_ui_scroll

@cjming: Thank you so much for verifying/clarifying!

So no further action is needed except, like I mentioned earlier, removing the sampling logic from the instrument.

oh whoops! you're right @mpopov --- i should read more carefully -- @KSarabia-WMF please disregard my earlier message - somehow i conflated the two.

FYI to folks, just to be super careful, we probably won't deploy this until after the data collection for T367871 is done.