Page MenuHomePhabricator

Database error while saving an article (1205 Lock wait exceeded in Title::invalidateCache)
Closed, ResolvedPublic


Article: Liste_der_denkmalgeschützten_Objekte_in_Wien/Währing
error message:
Es ist ein Datenbankabfragefehler aufgetreten. Dies könnte auf einen Fehler in der Software hindeuten.
Funktion: Title::invalidateCache
Fehler: 1205 Lock wait timeout exceeded; try restarting transaction (
But the edit looks good:

Event Timeline

Translation-service for non german unterstandig people:
Es ist ein Datenbankabfragefehler aufgetreten. Dies könnte auf einen Fehler in der Software hindeuten. => A database error occured. This may indicate a bug in the software.
Funktion => Function
Fehler => Error

@Boshomi Thanks for reporting, it seems that the edit was completed ok. Can you reproduce the error again, for that article or another?

Databases were in a poorer state for some hours after the datacenter failover. The actual infrastructure stability was not affected, but there could be some more errors than usual. On my logs, however, those errors were rare (50 for all users in 24 hours) amd those should not happen again in an eventual next failover. They also should not longer be happenning.

Could you confirm that this is true?

jcrespo triaged this task as Medium priority.
jcrespo moved this task from Triage to In progress on the DBA board.

@jcrespo no, I can not reproduce this bug.

What I did before:

I opened 60+ articles and mad a simple replace. then I saved the open tabs with Alt+S while changing the tabs with mouse. So I can save up to 3 articles per second.

one of this sessions ended at UTC 18:06:19; between 18:06:19 and 18:07 I mad the same simple edit in the last 27 artikles and started with saving some secondes before 18:08 the diff 153655394 was the first article in this session. The edit with the error message was as 18:07:55 the next edit followed at 18:00:00, I think, I saved the tabs much faster. I suspect a delay in messaging or database near by 18:00

on April 20 I startet edting with this edit at UTC 15:00
between 15:00 and 18:10 I mad about 900 edits, the change of the datacenter was no problem, the reported error message was unique.

Aklapper renamed this task from Database error while saving a artice to Database error while saving an article (1205 Lock wait exceeded in Title::invalidateCache).Apr 23 2016, 7:03 AM
jcrespo lowered the priority of this task from Medium to Low.

I am not sure that should be supported, Bot policy on ment says:

Bots running without a bot flag should edit at intervals of over 1 minute. Once they have been authorised and appropriately flagged, they should operate at an absolute minimum interval of 5 seconds (12 edits per minute)

You say you are editing 3 articles per second on your own.

And even you are not a bot, because you are simply using a javascripts on a browser, 3 edits per second may not be something it is supported, specially if not using the API (which has lag into account). I am not saying there is no issue here, but certainly it is not a common traffic pattern.

@jcrespo: I save 3 article per second, but the mean time for simplest edit is about 2 seconds. (yes that is faster than a bot, but I have no timer, I do this per hand)

try to open 60+ pages with chrome/chromium on linux, you can scroll with the mousewheel over the tabs => the browser switch from one tab to another.

hold Alt+E while scroll the mousewheel => all the tabs are in edit mode

faster that Alt+E is to open links like with Linkclump (Browser-Addon)

I make a simple edit, check the diff. When ready I save all the open tabs with Alt+S, normally with one keystroke per page. (holding Alt+S and scroll over the tab is potential faster, but I get a lot of edit conflicts by my self, and it is possible to run into the database limits which is beyond 10 edits per second. )

yes that is not a common traffic pattern. I simple reduced unproductive time to a minimum.

Marostegui subscribed.

I am going to close this as resolved as it looked a one time thing, and as specified here T133185#2230459 it could have been a cause of cold servers after the failover.
Also a lot has happened between the time when this task was opened and now, lots of schema changes, mariadb migration from 10.0 to 10.1 etc.

If someone feels this needs to be kept open, feel free to reopen!