Page MenuHomePhabricator

monuments_all table is missing
Closed, ResolvedPublic

Description

@JeanFred The monuments_all table doesn't exist right now!

Event Timeline

JeanFred claimed this task.
JeanFred raised the priority of this task from to Unbreak Now!.
JeanFred updated the task description. (Show Details)
JeanFred added subscribers: JeanFred, Emijrp, Multichill.

Curse! O_O

This is not supposed to happen.

Regenerating the table right now.

JeanFred renamed this task from monuments_all table is dropped to monuments_all table is missing.Sep 2 2015, 9:22 PM
JeanFred set Security to None.

There is a "monuments_all_tmp" table, btw.

Curse! O_O

This is not supposed to happen.

Regenerating the table right now.

Script succeeded. Should be back now. Can someone confirm?

There is a "monuments_all_tmp" table, btw.

Yes, that’s how fill_table_monuments_all.sql proceeds : merging all data in monuments_all_tmp, dropping monuments_all, renaming monuments_all_tmp to monuments_all. So something went wrong after dropping but before renaming, which is odd.

There is something wrong in monuments_all again. SQL queries are sloooow and don't finish. In _tmp are OK.

There is something wrong in monuments_all again. SQL queries are sloooow and don't finish. In _tmp are OK.

Yeah, I was noticing that too. I think it is because the categorisation job and the update table job were running at the same time. I killed the former and the latter finishes. How is it on your end @Emijrp?

Can you please update the database only handly?

I don't think that Wikipedia lists change a lot, so we don't get much updating it (daily?) with cron.

I think that it is better to have "old" data in the map, that break the database and have a map without a monument.

Can you please update the database only handly?

I don't think that Wikipedia lists change a lot, so we don't get much updating it (daily?) with cron.

I think that it is better to have "old" data in the map, that break the database and have a map without a monument.

It’s actually the other way around. In both incidents, I had started a manual update (following hot config changes) that turned out to be clashing with the long-running job − so, yeah, completely my fault, but obviously I won’t be doing that again.