All The Tropes Wiki Project, Part XXIV

[robkelk]

We might not need to ask for CommentStreams, if I'm reading two entries in the migration announcement correctly:

    Once the migration is completed, does that mean the Comments extension will finally be enabled on all wikis? ?️ my Name is Katsumi KamenRiderRevice-logo.webp talk KamenRiderRevice-logo.webp contributions 13:50, 8 January 2022 (UTC)
    The comments extension is active when connecting via SCSVG, yes. This will be automatic. ~ RhinosF1 - (chat)· acc· c - (    online) 12:10, 9 January 2022 (UTC)

Oh, and it's now 75 minutes into the 30-minute migration and Miraheze is still read-only. Not that I expected a seamless migration - from the announcement, it's obvious that they had no way to test things beforehand - but we can't even get a status update on Meta.

EDIT: No updates on their Twitter feed, either. And Phabricator is broken - https://phabricator.miraheze.org/ only displays error messages when Miraheze is in read-only mode. (Bad coding there, folks - the homepage should always be reachable, even if only to display a 404 message. What's happening now makes it look like the entire site is broken. somebody remind me to enter a Phabricator ticket about the bork once Phabricator is back online, please.)

Comments is for the blog comments. CommentStreams is a more wiki friendly discussion format without Flow issues and SocialProfile integration with avatars.

So we still want it, then. Thanks.

[Bob Schroeck]

It's almost 9 PM EST and according to Community Noticeboard the migration is done... but we're getting 502 bad gateway errors.

[robkelk]

I'm not surprised - the databases need to be rebuilt, and that takes processor power. (This isn't the first server migration I've been involved with this year.)

If we're still seeing the 502s and 503s this time tomorrow, then I'll be worried.

[Bob Schroeck]

Another thing: The patrolled bit on a lot of articles from the last couple days has been undone.

[robkelk]

Parser & Memcache should fill up once the page has been loaded for the first time. Once most pages have loaded at least once and well used templates have populated the cache then it will get better for any page that uses that template. RhinosF1 (Miraheze) (talk) 11:09, 14 January 2022 (UTC)

If RhinosF1 is saying what I think RhinosF1 is saying, we need to load most of our pages before the problems go away. Unless somebody wants to load all 157,327 article pages on ATT by hand (at five seconds per page, that's over 218 hours of non-stop work), we need a bot run to display every article page on the wiki. I've asked whether there's any way to automate this, since IIRC our bots don't display pages (they just edit pages). EDDIT: Correction: I'm trying to ask, but the editor keeps timing out.

[robkelk]

...
EDIT: No updates on their Twitter feed, either. And Phabricator is broken - https://phabricator.miraheze.org/ only displays error messages when Miraheze is in read-only mode. (Bad coding there, folks - the homepage should always be reachable, even if only to display a 404 message. What's happening now makes it look like the entire site is broken. somebody remind me to enter a Phabricator ticket about the bork once Phabricator is back online, please.)

RE-EDIT: 00:30 UTC and the wikis appear to be back... and slow, and tossing 502 and 503 errors up more than I'd prefer to see. As expected, the "recent changes" pages show only changes made on the new server, although the edit history is still present. Phabricator is still down.

And now it appears that we can't login to Phabricator. So this still needs to be reported... later.

[robkelk]

Looking at the Community noticeboard on Meta, I see that it's possible to subscribe to individual sections on that page. Maybe this might make for a good replacement for Flow? I don't know how their pages work behind the scenes, except that they use a "Parsoid/RESTBase server" (which at the moment keeps timing out... so maybe that isn't such a good replacement for Flow.)

[GethN7]

Looking at the Community noticeboard on Meta, I see that it's possible to subscribe to individual sections on that page. Maybe this might make for a good replacement for Flow? I don't know how their pages work behind the scenes, except that they use a "Parsoid/RESTBase server" (which at the moment keeps timing out... so maybe that isn't such a good replacement for Flow.)

Parsoid/RESTbase refers to the VisualEditor, which uses the aforementioned to turn wiki-text into client-side HTML (which looks nicer to search engines and automates certain aspects of wiki linking and text parsing done by hand in source mode). Ordinary editing (stock MediaWiki and/or CommentStreams which we plan to enable once the dust settles) uses the stock text parser of the MediaWiki engine, which does not use Parsoid/RESTbase.

The issues being sorted out are the fact all posts made with the VisualEditor put text in their own tables which needs to be rebuilt and synced to the new database like everything else.

Ergo, our plan to depreciate Flow ASAP for CommentStreams is still our best option, and VisualEditor is optional for all users who choose to use it, it can be disabled by user preference.

Performance is rebounding drastically, pages loads that took forever are now much shorter. As the database updates some more, this should even out further. Will submit the changes to our discussion format to Phabricator when it's up again.

[robkelk]

I know there's a trope in here, but I don't know what it's called... or even if we have it.

[Bob Schroeck]

I'm drawing a blank, but yeah, I agree -- there's a trope there.

[Black Aeronaut]

A trope about obnoxious tourists?

[Umbire]

There's definitely a trope about stealing the spotlight that I'm more than sure we have.

[GethN7]

I have filed the appropriate Phabricator request for the changes to our discussion pages here:

https://phabricator.miraheze.org/T8618

I recommend NOT trying to log in via Github, that has cURL errors, other login options are fine if you wish to add anything to this request.

[Bob Schroeck]

Just a heads up... ShawnTehLogoBoi is back from his month-long tempban. He's had a small edit in the mod queue since Saturday, which I've just okayed.

[Bob Schroeck]

Is anyone else getting a database error when they search on something that isn't a page title?

[robkelk]

I am, yes. Searching for "Foo Fu" returned the page, while searching for "Foo" returned

A database query error has occurred. This may indicate a bug in the software.
[7be2299940ad71c690989ee2] 2022-01-18 22:52:17: Fatal exception of type "Wikimedia\Rdbms\DBQueryError"

And I got the autocompleting menu when I searched for "Foo", too.

[GethN7]

Is anyone else getting a database error when they search on something that isn't a page title?

I can explain this. One of the biggest parts of a MediaWiki database is the searchindex tables, they can get MASSIVE, and given the sheer scale of the move that had been done, the tables are likely still rebuilding, and while most of the cache of valid pages is still present, they are rebuilding the search index to better optimize it for the new servers, so errors like this are to be expected. 

The good news is this not a permanent matter, and if they can get an ElasticSearch cluster to replace the bog standard searchindex cache, we will have a much faster and accurate searchindex table optimized for very high server loads.

[Bob Schroeck]

Okay, that I can cope with for now. Thanks.

[robkelk]

* Nightmare Fuel: TBA

Really?

That's even worse than "* Funny Moments: Bound to happen in each collection."

So, yeah, All The Tropes:CONTEXT now has a new entry that begins with the text quoted above and ends with "If you're going to write the example later, then write the entire example later; don't leave people hanging. "

[Bob Schroeck]

That's the one good thing about the people who might otherwise be more trouble than they're worth as contributors: at least they help us find the edge cases and unexpected situations that need to be addressed explicitly in our policies and guidelines.

[Bob Schroeck]

Is anyone else getting a database error when they search on something that isn't a page title?

I can explain this. One of the biggest parts of a MediaWiki database is the searchindex tables, they can get MASSIVE, and given the sheer scale of the move that had been done, the tables are likely still rebuilding, and while most of the cache of valid pages is still present, they are rebuilding the search index to better optimize it for the new servers, so errors like this are to be expected. 

The good news is this not a permanent matter, and if they can get an ElasticSearch cluster to replace the bog standard searchindex cache, we will have a much faster and accurate searchindex table optimized for very high server loads.

Just as a note to everyone, the indexes are apparently still rebuilding.

[robkelk]

I would not be surprised if they continue to rebuild for the rest of the week. ATT is the largest wiki at Miraheze, after all - we're going to take longer than anybody else.

[Bob Schroeck]

Yeah, that's going to be annoying.

[Umbire]

* Nightmare Fuel: TBA

Really?

That's even worse than "* Funny Moments: Bound to happen in each collection."

So, yeah, All The Tropes:CONTEXT now has a new entry that begins with the text quoted above and ends with "If you're going to write the example later, then write the entire example later; don't leave people hanging. "

Not gonna lie, I've been guilty of this at least once (in fairness, I was literally about to add the example in question right after - I've had issues with letting WIP edits sit and then typically losing them).

[Bob Schroeck]

This was someone who's already undergone a couple tempbans and has had more than a few edits rejected for generally stupid stuff -- this was pretty much par for the course with them.