Page MenuHomePhabricator

lab's templatelinks contain possible garbage rows
Closed, ResolvedPublic

Description

I noticed templatelinks table on labs contain garbage rows which I can't explain where they come from, for example:

use hewiki_p;
select * from templatelinks where tl_from=1286698 and tl_namespace=10 and tl_title='בעבודה';
+---------+--------------+--------------+-------------------+
| tl_from | tl_namespace | tl_title     | tl_from_namespace |
+---------+--------------+--------------+-------------------+
| 1286698 |           10 | בעבודה       |                 0 |
+---------+--------------+--------------+-------------------+

So page with id 1286698 (https://he.wikipedia.org/w/index.php?title=%D7%A7%D7%95%D7%A0%D7%92_%D7%A4%D7%99%D7%95%D7%A8%D7%99) transcludes "בעבודה" template. Looking on the page itself - it doesn't. The last time such template was on this page was 12 June 2015 (https://he.wikipedia.org/w/index.php?title=%D7%A7%D7%95%D7%A0%D7%92_%D7%A4%D7%99%D7%95%D7%A8%D7%99&diff=17074119&oldid=17062970).

Event Timeline

eranroz created this task.Jan 28 2017, 7:50 PM
Restricted Application added a subscriber: Aklapper. · View Herald TranscriptJan 28 2017, 7:50 PM
jcrespo closed this task as Resolved.Jan 28 2017, 8:00 PM
jcrespo claimed this task.
jcrespo added a subscriber: jcrespo.

The main question here is if this is a labs issue or a production issue shown on labs. If it is production, that is more or less "normal", on a next pass of a maintenance script it will be gone.

If it differs from production -it is a labs-only issue-, most of these problems are known, and all of these will be solved/marked invalid when the new labsdb servers, with re-synced data, are fully available.

From a quick look, it looks like the second.

localhost[hewiki_p]> select * from templatelinks where tl_from=1286698 and tl_namespace=10 and tl_title='בעבודה';
+---------+--------------+--------------+-------------------+
| tl_from | tl_namespace | tl_title     | tl_from_namespace |
+---------+--------------+--------------+-------------------+
| 1286698 |           10 | בעבודה       |                 0 |
+---------+--------------+--------------+-------------------+
1 row in set (0.01 sec)

localhost[hewiki_p]> delete from hewiki.templatelinks where tl_from=1286698 and tl_namespace=10 and tl_title='בעבודה';
Query OK, 1 row affected (0.00 sec)

localhost[hewiki_p]> select * from templatelinks where tl_from=1286698 and tl_namespace=10 and tl_title='בעבודה';
Empty set (0.00 sec)

I have deleted manually the row- but sadly this can happen again, here it is why: https://wikitech.wikimedia.org/wiki/Help:Tool_Labs/Database/Replica_drift Only the new setup with the new servers will fix this permanently.