On that note, Jonathan, do you have SQL queries that return the same results as the Wikimetrics reports?  I tried writing my own for bytes added, but it's pretty janky and takes forever to return anything in Quarry.  Would you share your wisdom with a poor wayward amateur?


On Tue, Feb 17, 2015 at 9:35 AM, Amanda Bittaker <abittaker@wikimedia.org> wrote:
Sweet, thanks Jonathan.  I added a "Heartbreak" token, because at this point I am really far too emotionally attached to Wikimetrics.  

On Tue, Feb 17, 2015 at 9:25 AM, Jonathan Morgan <jmorgan@wikimedia.org> wrote:
Hi Amanda,

Here's a ticket you can upvote: https://phabricator.wikimedia.org/T87596

I added a link to this thread to the task. I also added an "Evil Spooky Haunted Tree" token to the task. Because... well it just felt like the right thing to do.

- J 

On Tue, Feb 17, 2015 at 7:58 AM, Dan Andreescu <dandreescu@wikimedia.org> wrote:
I can't find a specific ticket, Nuria may know of one.  In general, this is the project that LabsDB tickets are tagged with: https://phabricator.wikimedia.org/tag/wikimedia-labs-infrastructure/

On Tue, Feb 17, 2015 at 10:53 AM, Amanda Bittaker <abittaker@wikimedia.org> wrote:
Good morning Dan,

Thanks very much for the explanation.  Is there a Phabricator task we can upvote (award a token?) to make this issue more visible?  

As always, we really appreciate your help with this.

Best,
Amanda



On Tue, Feb 17, 2015 at 7:20 AM, Dan Andreescu <dandreescu@wikimedia.org> wrote:
Sorry for the trouble, Amanda.  The problem is solely with the underlying database, which we don't maintain.  It's a sanitized replica of all the changes being made to all the wikis so it's a fairly complicated piece of infrastructure that sometimes has problems.  The folks who maintain it are aware of the issues, but we'll continue representing them until they're solved.

On Mon, Feb 16, 2015 at 3:49 PM, Amanda Bittaker <abittaker@wikimedia.org> wrote:
Oop, thanks for the ping, Nuria.  Wikimetrics seems to be working better now.  I still get failures, especially when running three or four reports in one batch, but the reports work if you rerun them (sometimes a couple times.)

I'm still getting "PENDING"s that turn into "FAILURE"s sometimes, which I just noticed for the first time last Thursday.  Also, sometimes the "FAILURE"s change position in the Current Report Inbox list, moving up or down a spot.  Not sure if that helps diagnose what might be happening...

In any case, Wikimetrics is mostly functioning but seems to be having recurring troubles that sometimes blow up to freeze the whole tool.  It would be great to resolve the troubles before the next explosion--is there anything I can do to help?  Dan H and I still have plenty of reports to run, we can keep you updated on the reports ran and failure rate while you are fixing, if that would be useful.

Many thanks,
Amanda


On Mon, Feb 16, 2015 at 10:15 AM, Nuria Ruiz <nuria@wikimedia.org> wrote:
Ping ....

On Fri, Feb 13, 2015 at 2:19 PM, Nuria Ruiz <nuria@wikimedia.org> wrote:
Amanda, 

Looks like wikimetrics was able to run automatic reports last night w/o big issues, are your reports still failing?

Thanks, 

Nuria

On Thu, Feb 12, 2015 at 1:42 PM, Amanda Bittaker <abittaker@wikimedia.org> wrote:
Alright, thanks so much for your help once again, Nuria.

If there's anything I can do or any information I can contribute, please don't hesitate to ping me.

Best,
Amanda



On Thu, Feb 12, 2015 at 1:36 PM, Nuria Ruiz <nuria@wikimedia.org> wrote:
DB connections in labs look to be failing, unfortunately  I think besides asking for help on the labs list there is not much we can do there. I will start a thread on this regard.

Thanks, 

Nuria

On Thu, Feb 12, 2015 at 1:32 PM, Amanda Bittaker <abittaker@wikimedia.org> wrote:
Thanks so much for the quick response, Nuria.

I ran the exact same reports on the same cohort as one of the last batches that were failing.  Last time 2/4 of the reports failed, when I reran the individually they succeeded.  (But they don't always, I reran one report 3 times this morning before it worked.)   This time, my failure rate got worse:  4/4 failed, although they said "PENDING" for a few seconds first, which is new.

Is that useful information?  Please do let me know what else I can do to help solve this.

Thanks again,
Amanda



On Thu, Feb 12, 2015 at 1:09 PM, Jonathan Morgan <jmorgan@wikimedia.org> wrote:
Thanks Nuria!

On Thu, Feb 12, 2015 at 12:57 PM, Nuria Ruiz <nuria@wikimedia.org> wrote:

If so a cohort + report to repro will be most useful.

Translation:* try to run the exact same reports on the same cohort again, to see if the same metrics fail. Let us know what you find. ;)

Same goes for anyone else who experiences these issues: the more details we (users) can provide the engineers, the more effective they can be at diagnosing and addressing the problems. 

Cheers,
- J

*for anyone who is not 100% familiar with that hip, new software engineering lingo



Thanks, 

Nuria




On Thu, Feb 12, 2015 at 12:35 PM, Dan Andreescu <dandreescu@wikimedia.org> wrote:
Recently there was a restart of the labsdb cluster.  I'm sorry but I don't have time to check on it, but I bet that's the problem.  I'm off tomorrow unfortunately but I'll try to check tomorrow night :(  I hope someone else beats me to it.

On Thu, Feb 12, 2015 at 3:20 PM, Jonathan Morgan <jmorgan@wikimedia.org> wrote:
(ping Kevin and Dan A.)

Hi Amanda, I've had some problems with report failures recently when I ran a few test cohorts. On the same cohort, when I ran multiple concurrent reports (say, bytes added, edits, and pages created), some would fail and others succeed. It wasn't clear what the issue was. 

- J

On Thu, Feb 12, 2015 at 12:16 PM, Amanda Bittaker <abittaker@wikimedia.org> wrote:
Hello all,

I am getting failures again, both when uploading cohorts and running reports.  Strangely, it seems the more reports you try to run in one batch the less likely it is any report will succeed.

Is anyone else having these problems again?  Wonderful Analytics people, could you please work your magic again?

Many thanks,
Amanda

_______________________________________________
Wikimetrics mailing list
Wikimetrics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikimetrics




--
Jonathan T. Morgan
Community Research Lead
Wikimedia Foundation



_______________________________________________
Wikimetrics mailing list
Wikimetrics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikimetrics





--
Jonathan T. Morgan
Community Research Lead
Wikimedia Foundation


_______________________________________________
Wikimetrics mailing list
Wikimetrics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikimetrics








_______________________________________________
Wikimetrics mailing list
Wikimetrics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikimetrics



_______________________________________________
Wikimetrics mailing list
Wikimetrics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikimetrics



_______________________________________________
Wikimetrics mailing list
Wikimetrics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikimetrics



_______________________________________________
Wikimetrics mailing list
Wikimetrics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikimetrics




--
Jonathan T. Morgan
Community Research Lead
Wikimedia Foundation


_______________________________________________
Wikimetrics mailing list
Wikimetrics@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/wikimetrics