Page MenuHomePhabricator

Description page sometimes not created during upload
Open, MediumPublic

Description

Splitting off from T32551.

Occasionally an upload succeeds without a description page being created.

Its unclear what circumstances this happens. Possibly an error connecting to externalstorage causes an exception to be thrown, or something of that nature.

There is a list at https://commons.wikimedia.org/wiki/User:Ilmari_Karonen/Queries/Zombie_images

As of this writing, some examples: File:Jersey, Corbiere Lighthouse by moonlight, Channel Islands-LCCN2002696513.tif, File:Palais de Justice and bridge to exchange, Paris, France-LCCN2001698506.jpg

File:Château du Martinet - façade nord.JPG is missing log entries, but the page is fine.

Most of the examples seem to be from mass upload tools (GWToolset and flickr2commons). I expect that to just be coincidence though

Consider bug T32551 to be about the referential integrity issues where people couldn't create the pages due to missing revisions, and this to be about the larger issue where sometimes pages aren't created

Details

Reference
bz66355

Event Timeline

bzimport raised the priority of this task from to Medium.Nov 22 2014, 3:10 AM
bzimport set Reference to bz66355.
bzimport added a subscriber: Unknown Object (MLST).
Steinsplitter set Security to None.
Steinsplitter moved this task from Incoming to Uploading on the Commons board.
Bawolff raised the priority of this task from Medium to High.Jun 1 2015, 2:38 PM

I'm raising priority, as there was a significant uptick in this happening over the past few days. See https://commons.wikimedia.org/w/index.php?title=User%3AIlmari_Karonen%2FQueries%2FZombie_images&type=revision&diff=162295400&oldid=155762633

Tgr lowered the priority of this task from High to Medium.Jun 22 2015, 9:33 PM

I'm raising priority, as there was a significant uptick in this happening over the past few days. See https://commons.wikimedia.org/w/index.php?title=User%3AIlmari_Karonen%2FQueries%2FZombie_images&type=revision&diff=162295400&oldid=155762633

It looks like whatever caused that spike got fixed (or maybe it was caused by a specific upload tool in which case that tool should be identified):

mysql:research@analytics-store.eqiad.wmnet [commonswiki]> select substr(img_timestamp, 1, 8), count(*) from image left join page on page_namespace = 6 and page_title = img_name where page_title is null group by substr(img_timestamp, 1, 8) order by substr(img_timestamp, 1, 8);
+-----------------------------+----------+
| substr(img_timestamp, 1, 8) | count(*) |
+-----------------------------+----------+
|                             |        1 |
| 20051120                    |        1 |
| 20090829                    |        1 |
| 20140510                    |        1 |
| 20140527                    |        1 |
| 20140603                    |        1 |
| 20140606                    |        4 |
| 20140609                    |        1 |
| 20140611                    |        2 |
| 20140620                    |        1 |
| 20140924                    |        1 |
| 20141008                    |        1 |
| 20150310                    |        1 |
| 20150314                    |        1 |
| 20150318                    |        1 |
| 20150320                    |        1 |
| 20150321                    |        1 |
| 20150326                    |        1 |
| 20150329                    |        1 |
| 20150331                    |        1 |
| 20150402                    |        6 |
| 20150403                    |        1 |
| 20150424                    |        4 |
| 20150425                    |        1 |
| 20150427                    |        1 |
| 20150430                    |        7 |
| 20150503                    |        1 |
| 20150515                    |        1 |
| 20150518                    |        1 |
| 20150523                    |        1 |
| 20150527                    |        7 |
| 20150528                    |        4 |
| 20150529                    |       16 |
| 20150530                    |       90 |
| 20150531                    |        2 |
| 20150601                    |       27 |
| 20150605                    |        1 |
| 20150609                    |        1 |
+-----------------------------+----------+
38 rows in set (5 min 55.52 sec)
Steinsplitter raised the priority of this task from Medium to High.
Steinsplitter added subscribers: Aklapper, Wilfredor.

Most of the examples seem to be from gwtoolset.

But also flickr2commons. Generally massuploads. Seems not GWT only related.

Steinsplitter updated the task description. (Show Details)
Steinsplitter updated the task description. (Show Details)

Almost all of the bad uploads happened between 2015-05-26 and 2015-06-02, ie. the period when Commons was on 1.26wmf7. So probably something in that branch made uploads more fragile, but got fixed by next week.

Tgr lowered the priority of this task from High to Medium.Jun 30 2015, 7:57 PM

Lowering priority as the last occurrence of this bug was three weeks ago.

(diff | hist) . . User:Ilmari Karonen/Queries/Zombie images‎; 12:18 . . (-1,424)‎ . . ‎Dispenser (talk | contribs | block)‎ (Wikimedia Foundation: Where the lost edits don't matter) [rollback] [br]


Anything new here? Looks like users are unhappy because of this bug.