cjming opened https://gitlab.wikimedia.org/repos/data-engineering/mpic/-/merge_requests/211

Update instruments:

hi @phuedx @Sfaci - as I dig into the JS client and root around in the EventLogging extension, I'm deducing that the MetricsClient used by Instrument is instantiated by initMetricsClient() which calls newMetricsClient() with streams configs from the EventLogging hook getModuleData which calls getEventLoggingConfig() which calls loadEventStreamConfigs() which calls the EventLogging.StreamConfigs service which in turn gets stream configs from the EventStreamConfig.StreamConfigs service which gets stream configs from StreamConfigsFactory.

It looks like in StreamConfigsFactory::getInstance():

public function getInstance(): StreamConfigs {
	$streamConfigs = [];
	$this->hookRunner->onGetStreamConfigs( $streamConfigs );

	$streamConfigs = array_merge(
		$streamConfigs,
		$this->options->get( 'EventStreams' )
	);

	return new StreamConfigs(
		$streamConfigs,
		$this->options->get( 'EventStreamsDefaultSettings' ),
		$this->logger
	);
}

the array_merge() function overrides whatever is provided by the onGetStreamConfigs() hook (InstrumentConfigsFetcher) with what is set by static config.

Here's my stream config provided by my local xLab (where I selected contextual attributes) before array_merge:

And here are stream configs after array_merge (note overridden numeric keys):

For reference, here is my local producer config which closely matches prod config (minus the mediawiki attributes):

'product_metrics.web_base' => [
	'schema_title' => 'analytics/product_metrics/web/base',
	'producers' => [
		'metrics_platform_client' => [
			'provide_values' => [
				'agent_client_platform',
				'agent_client_platform_family',
				'performer_is_logged_in',
				'performer_is_temp',
				'performer_pageview_id',
				'performer_active_browsing_session_token',
			],
		],
	],
],

When I was testing this locally, I got events validated for a new PageVisit instrument:

const pageVisitInstrument = mw.eventLog.newInstrument(
	"page-visits-demo",
	"product_metrics.web_base",
	"/analytics/product_metrics/web/base/1.4.2"
);

pageVisitInstrument.submitInteraction(
	"page-visited",
	{ action_context: 'test instrument via xLab'}
);

But as confirmed by the debugging screenshots above, I only got the contextual attributes with the event that were defined by my local $wgEventStreams config:

Actually - I didn't get performer_is_temp - not sure why - need to investigate that more.

Anyway, my question is what is the best approach for making xLab config supercede hard-coded config? IIRC there was some discussion about hard-coded config should always take precedence? Is this is still the case? If an instrument owner wants to override the default contextual attributes set for product_metrics.web_base, do we want to allow for this?

sfaci merged https://gitlab.wikimedia.org/repos/data-engineering/mpic/-/merge_requests/211

Update instruments endpoint and contextual attributes

Moving this ticket to blocked until all tasks related to T398480: [Epic] Update instruments configuration for xLab are resolved.

Change #1167322 had a related patch set uploaded (by Clare Ming; author: Clare Ming):

[operations/deployment-charts@master] xLab: Deploy v0.7.8 release to staging

https://gerrit.wikimedia.org/r/1167322

Change #1167322 merged by jenkins-bot:

[operations/deployment-charts@master] xLab: Deploy v0.7.8 release to staging

https://gerrit.wikimedia.org/r/1167322

Change #1171319 had a related patch set uploaded (by Clare Ming; author: Clare Ming):

[operations/deployment-charts@master] xLab: Deploy v0.7.9 release to staging

https://gerrit.wikimedia.org/r/1171319

Change #1171319 merged by jenkins-bot:

[operations/deployment-charts@master] xLab: Deploy v0.7.9 release to staging

https://gerrit.wikimedia.org/r/1171319