If I were to venture into writing a changeset for this (made into a task: https://phabricator.wikimedia.org/T87177 ), is everything self-contained in the EventLogging extension or are there external parts involved in the current pipeline sending events to the DB in production that I need to be aware of?
On Fri, Jan 9, 2015 at 8:40 AM, Gilles Dubuc <[email protected]> wrote: > I think Gilles and Erik want to calculate page views for GLAM mainly >> (although there are some other good reasons too) -- sampling would probably >> be ok but we'd miss the long tail of views. >> > > That's correct. We're looking to compile media view counts as accurate as > the ones we have for article views at the moment. Sampling would be fine to > identify the X most viewed media across a wiki, but it definitely wouldn't > help small GLAMs who want to get that information about their own > collection, if their media happen to be "low traffic" in the grand scheme > of things. I think that the latter is the main use case for doing this, > which is why I'm looking for a solution that wouldn't involve sampling. > > Compiling the top list has entertainment value, letting GLAM contributors > get accurate statistics about their content improves the chances that they > will keep contributing more. I think that's more valuable than the > entertainment factor of the top list. > > On Wed, Jan 7, 2015 at 8:02 PM, Toby Negrin <[email protected]> wrote: > >> I think Gilles and Erik want to calculate page views for GLAM mainly >> (although there are some other good reasons too) -- sampling would probably >> be ok but we'd miss the long tail of views. >> >> On Wed, Jan 7, 2015 at 10:56 AM, Nuria Ruiz <[email protected]> wrote: >> >>> I see. My main point was that -regardless of collection method- we might >>> not need every single data point to calculate uniques. >>> >>> On Wed, Jan 7, 2015 at 10:38 AM, Toby Negrin <[email protected]> >>> wrote: >>> >>>> Yes -- we disabled it because there wasn't a use case. We have one now >>>> :) >>>> >>>> On Wed, Jan 7, 2015 at 10:32 AM, Nuria Ruiz <[email protected]> >>>> wrote: >>>> >>>>> > I believe there is already an EL-Kafka pipeline and this would make >>>>> it easy to integrate page views with our regular processing. >>>>> >>>>> Note that the pipeline was disabled 6 months ago and thus my comment >>>>> "in the near term" >>>>> >>>>> https://github.com/wikimedia/operations-puppet/commit/f85b1dbcd61bbb58684ff93704c1804e808a5d6e >>>>> >>>>> On Wed, Jan 7, 2015 at 9:39 AM, Toby Negrin <[email protected]> >>>>> wrote: >>>>> >>>>>> I'd also like us to consider routing this dataset to hadoop. I >>>>>> believe there is already an EL-Kafka pipeline and this would make it easy >>>>>> to integrate page views with our regular processing. >>>>>> >>>>>> Gilles -- are mobile page views included in your stream? >>>>>> >>>>>> -Toby >>>>>> >>>>>> On Wed, Jan 7, 2015 at 9:27 AM, Nuria Ruiz <[email protected]> >>>>>> wrote: >>>>>> >>>>>>> >Great, then I guess it's a matter of only making the data go to >>>>>>> files and not to DB for the particular schema we'll create. Does >that >>>>>>> sound like something feasible? How much work would be required to set >>>>>>> it up? >>>>>>> I do not think this is feasible on the near term w/o changes in our >>>>>>> end. I also am not sure it is really needed. You are concern about >>>>>>> sending >>>>>>> stuff to db due to "volume", correct? I do not understand why logging >>>>>>> every >>>>>>> single data point would be needed. Maybe you can explain that with a bit >>>>>>> more detail for us to grasp the use case? >>>>>>> >>>>>>> If it is a matter of identifying distinct requests that can be done >>>>>>> having sampled your dataset if it is large enough, we can help with that >>>>>>> and leila just put together some docs on this regard, while this is for >>>>>>> hive queries principles can apply elsewhere: >>>>>>> https://wikitech.wikimedia.org/wiki/Analytics/Cluster/Hive/Counting_uniques >>>>>>> >>>>>>> >>>>>>> >>>>>>> On Wed, Jan 7, 2015 at 6:42 AM, Gilles Dubuc <[email protected]> >>>>>>> wrote: >>>>>>> >>>>>>>> Right -- couldn't we just tag the URL? >>>>>>>>> >>>>>>>> >>>>>>>> The event of the user actually viewing the image is completely >>>>>>>> disconnected from the URL hit in Media Viewer, which is why we need EL >>>>>>>> and >>>>>>>> can't rely on existing server logs. >>>>>>>> >>>>>>>> >>>>>>>>> Eventlogging data currently does go to files, as well as to the DB. >>>>>>>>> >>>>>>>> >>>>>>>> Great, then I guess it's a matter of only making the data go to >>>>>>>> files and not to DB for the particular schema we'll create. Does that >>>>>>>> sound >>>>>>>> like something feasible? How much work would be required to set it up? >>>>>>>> >>>>>>>> On Tue, Jan 6, 2015 at 7:45 PM, Andrew Otto <[email protected]> >>>>>>>> wrote: >>>>>>>> >>>>>>>>> Eventlogging data currently does go to files, as well as to the >>>>>>>>> DB. Check it out on stat1003 at /srv/eventlogging/archive. >>>>>>>>> >>>>>>>>> If you need something with higher throughput then eventlogging >>>>>>>>> itself supports…then let’s talk :D >>>>>>>>> >>>>>>>>> -Ao >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> On Jan 6, 2015, at 13:28, Erik Zachte <[email protected]> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> You mean attach an X-analytics parameter, for extra images beyond >>>>>>>>> the one the user initially requested. >>>>>>>>> >>>>>>>>> But then we would undercount, basically missing all image views >>>>>>>>> from clicking right arrow in image viewer. >>>>>>>>> I'm not sure how much we would miss then. >>>>>>>>> iirc Gilles said this browsing feature was used quite a long, but >>>>>>>>> I'm not sure. >>>>>>>>> >>>>>>>>> >>>>>>>>> *From:* [email protected] [ >>>>>>>>> mailto:[email protected] >>>>>>>>> <[email protected]>] *On Behalf Of *Toby >>>>>>>>> Negrin >>>>>>>>> *Sent:* Tuesday, January 06, 2015 19:16 >>>>>>>>> *To:* A mailing list for the Analytics Team at WMF and everybody >>>>>>>>> who has an interest in Wikipedia and analytics. >>>>>>>>> *Subject:* Re: [Analytics] Making EventLogging output to a log >>>>>>>>> file instead of the DB >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> Right -- couldn't we just tag the URL? >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> On Tue, Jan 6, 2015 at 10:10 AM, Erik Zachte < >>>>>>>>> [email protected]> wrote: >>>>>>>>> >>>>>>>>> Just to clarify, this is about prefetched images which have not >>>>>>>>> been shown to the public. >>>>>>>>> >>>>>>>>> They were sent to the browser ahead of a possible request to speed >>>>>>>>> things up but in many cases never actually requested. >>>>>>>>> >>>>>>>>> >>>>>>>>> https://www.mediawiki.org/wiki/Requests_for_comment/Media_file_request_counts#Prefetched_images >>>>>>>>> >>>>>>>>> - Erik >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> *From:* [email protected] [mailto: >>>>>>>>> [email protected]] *On Behalf Of *Toby Negrin >>>>>>>>> *Sent:* Tuesday, January 06, 2015 18:49 >>>>>>>>> *To:* A mailing list for the Analytics Team at WMF and everybody >>>>>>>>> who has an interest in Wikipedia and analytics. >>>>>>>>> *Subject:* Re: [Analytics] Making EventLogging output to a log >>>>>>>>> file instead of the DB >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> Hi Gilles -- why won't the page view logs work by themselves for >>>>>>>>> this purpose? EL can be configured to write into Hadoop which is >>>>>>>>> probably >>>>>>>>> the best way to get the throughput you need but it seems >>>>>>>>> overcomplicated. >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> -Toby >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> On Tue, Jan 6, 2015 at 9:41 AM, Gilles Dubuc <[email protected]> >>>>>>>>> wrote: >>>>>>>>> >>>>>>>>> This depends on [1] so we're not going to need that immediately, >>>>>>>>> but in order to help Erik Zachte with his RfC [2] to track unique >>>>>>>>> media >>>>>>>>> views in Media Viewer, I'm going to need to use something almost >>>>>>>>> exactly >>>>>>>>> like EventLogging. The main difference being that it should skip >>>>>>>>> writing to >>>>>>>>> the database and write to a log file instead. >>>>>>>>> >>>>>>>>> That's because we'll be recording around 20-25M image views per >>>>>>>>> day, which would needlessly overload EventLogging for little purpose >>>>>>>>> since >>>>>>>>> the data will be used for offline stats generation and doesn't need >>>>>>>>> to be >>>>>>>>> made available in a relational database. Of course if storage space >>>>>>>>> and >>>>>>>>> EventLogging capacity were no object, we could just use EL and keep >>>>>>>>> the >>>>>>>>> ever-growing table forever, but I have the impression that we want to >>>>>>>>> be >>>>>>>>> reasonable here and only write to a log, since that's what Erik needs. >>>>>>>>> >>>>>>>>> So here's the question: for a specific schema, can EventLogging >>>>>>>>> work the way it does but only record hits to a log file (maybe it >>>>>>>>> already >>>>>>>>> does that before hitting the DB?) and not write to the DB? If not, how >>>>>>>>> difficult would it be to make EL capable of doing that? >>>>>>>>> >>>>>>>>> >>>>>>>>> [1] https://phabricator.wikimedia.org/T44815 >>>>>>>>> [2] >>>>>>>>> https://www.mediawiki.org/wiki/Requests_for_comment/Media_file_request_counts >>>>>>>>> >>>>>>>>> >>>>>>>>> _______________________________________________ >>>>>>>>> Analytics mailing list >>>>>>>>> [email protected] >>>>>>>>> https://lists.wikimedia.org/mailman/listinfo/analytics >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> _______________________________________________ >>>>>>>>> Analytics mailing list >>>>>>>>> [email protected] >>>>>>>>> https://lists.wikimedia.org/mailman/listinfo/analytics >>>>>>>>> >>>>>>>>> >>>>>>>>> _______________________________________________ >>>>>>>>> Analytics mailing list >>>>>>>>> [email protected] >>>>>>>>> https://lists.wikimedia.org/mailman/listinfo/analytics >>>>>>>>> >>>>>>>>> >>>>>>>>> >>>>>>>>> _______________________________________________ >>>>>>>>> Analytics mailing list >>>>>>>>> [email protected] >>>>>>>>> https://lists.wikimedia.org/mailman/listinfo/analytics >>>>>>>>> >>>>>>>>> >>>>>>>> >>>>>>>> _______________________________________________ >>>>>>>> Analytics mailing list >>>>>>>> [email protected] >>>>>>>> https://lists.wikimedia.org/mailman/listinfo/analytics >>>>>>>> >>>>>>>> >>>>>>> >>>>>>> _______________________________________________ >>>>>>> Analytics mailing list >>>>>>> [email protected] >>>>>>> https://lists.wikimedia.org/mailman/listinfo/analytics >>>>>>> >>>>>>> >>>>>> >>>>>> _______________________________________________ >>>>>> Analytics mailing list >>>>>> [email protected] >>>>>> https://lists.wikimedia.org/mailman/listinfo/analytics >>>>>> >>>>>> >>>>> >>>>> _______________________________________________ >>>>> Analytics mailing list >>>>> [email protected] >>>>> https://lists.wikimedia.org/mailman/listinfo/analytics >>>>> >>>>> >>>> >>>> _______________________________________________ >>>> Analytics mailing list >>>> [email protected] >>>> https://lists.wikimedia.org/mailman/listinfo/analytics >>>> >>>> >>> >>> _______________________________________________ >>> Analytics mailing list >>> [email protected] >>> https://lists.wikimedia.org/mailman/listinfo/analytics >>> >>> >> >> _______________________________________________ >> Analytics mailing list >> [email protected] >> https://lists.wikimedia.org/mailman/listinfo/analytics >> >> >
_______________________________________________ Analytics mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/analytics
