Page MenuHomePhabricator

Request creation of wikisource VPS project
Closed, ResolvedPublic

Description

Project Name: wikisource

Wikitech Usernames of requestors: MusikAnimal, Tpt, Samwilson, Mooeypoo, HMonroy, Dom Walden, Aezell

Purpose: As part of the #1 wish of the 2020 Wishlist Survey, and the #4 wish for 2019, we need a more robust environment to host the Wikisource Export tool. On Toolforge it has experienced frequent downtime. As an experiment we already moved the tool to a VPS instance under the commtech project, and the results are promising. However we'd like a dedicated project so that we can add volunteers and a staging instance.

Brief description: Most everything comes with Debian Buster, apparently. We will need to install Calibre (GNU General Public License). As for quota, I expect overall we won't need more than one m1.large and one m1.medium.

How soon you are hoping this can be fulfilled: Not urgent. Within a few weeks time would be preferable.

Event Timeline

can you point to the tasks around reliability issues with this on toolforge? is it due to anything inherent with the toolforge infrastructure?

can you point to the tasks around reliability issues with this on toolforge? is it due to anything inherent with the toolforge infrastructure?

T219330, T236391, and T242760, to name a few, and there are probably several relevant issues at https://github.com/wsexport/tool/issues.

I can't pinpoint exactly what about Toolforge is the issue, but I'm fairly confident it's part of it. We need to do more engineering, such as introducing a job queue. This is still doable on Toolforge but experience has shown that for high-traffic, resource-intensive tools like this, VPS is the better platform. We have many VPS tools as it is and are experienced with maintaining our own instances.

can you point to the tasks around reliability issues with this on toolforge? is it due to anything inherent with the toolforge infrastructure?

There have been various tasks over the years. One of the historical problems the tool had was leaking temp files on grid engine lighttpd nodes. Another was related to headless X server usage leaving orphan processes on the grid.

I think there has been a lot of work done to the tool that has helped in the past months, but this certainly is one of the tools that has usage to justify "graduation" from Toolforge to a custom project if that is desired. wsexport is linked to by the sidebar of https://en.wikisource.org/ for all users including anons as the means to download EPUB and MOBI versions of books.

I would certainly rather see it in its own project than stuffed into the commtech project. Grab bag projects are an anti-pattern to avoid. They make community involvement difficult and instance ownership difficult to determine.

MusikAnimal renamed this task from Request creation of wsexport VPS project to Request creation of wikisource VPS project.Feb 5 2020, 12:13 AM
MusikAnimal updated the task description. (Show Details)

Thanks, we concur this is deserving a dedicated project. Note I just changed the requested name to wikisource, in the off-chance other related tools get bundled into the project (we also have reservations about some connotations of wsexport :)

The wikisource CloudVPS project has been approved and created.

@MusikAnimal @Tpt @Samwilson @Mooeypoo @HMonroy @dom_walden @aezell,
You're all members of the new wikisource project. Please make sure you're subscribed to the cloud-annonce [0] mailing list so you can receive critical notifications regarding the CloudVPS service.

[0] https://lists.wikimedia.org/mailman/listinfo/cloud-announce