AO3 News

Post Header

Published:
2011-01-23 22:25:58 UTC
Translations:
Tags:

Thanks to the wonderful generosity of fandom, we've been able to buy a beautiful new set of servers! Meet the machines below!

The DB Server - aka "The Beast"

The Beast: cartoon style image of server

Our new database server is a serious badass. Built for us by open-source server specialists iXSystems, the DB server is an iX1204 powered by 2 latest-generation Intel 6-Core X5650 processors, and 48GB of RAM. This server will also have 2 Intel X25-M 80GB SSDs (solid-state drives, which have no moving parts) to make reads off the database crazy fast.

The Rails Twins

The Rails Twins: cartoon style image of servers

A set of identical twins, our speedy and strong new Rails servers will split the load of running the Archive app. These two are from the same family as the DB server: iX1204 models, each with 2x Intel Quad Core E5620 processors, 24GB of RAM, and 4x146GB 15k RPM hard drives.

The Originals

Front end server: cartoon style image of server

Slave: cartoon style image of server

Our trusty veterans! The original servers are another set of identical twins: Hewlett Packard ProLiant DL360 G5 models each powered by 2 Quad-Core Intel Xeon E5420 processors. They each also have 16GB RAM, and 4×72GB 15k RPM hard drives. For the last year, these two have had to carry the entire load all by themselves, but now they will both be moving into new roles:

  1. One of these machines will now serve as our friendly memcache server and it runs background and other odd jobs.

  2. The other one will become the DB's slave. *koff* (Of course the more vanilla-minded among us can think of it as being the DB's loyal sidekick!) It will also run our search engine.

The Storage Server

Storage server: cartoon style image of server

The QNAP TS-809U-RP is a 2U rackmount Network Attached Storage (NAS) appliance, intended for what the manufacturer describes as "massive data sharing" -- mmmhmm! We're loading this baby to capacity, with 16TB (8x2TB Hitachi Deskstar 7200RPM HDDs) in hot-swap bays to give us greatly increased data storage/backup capability.

The Switch

Switch: cartoon style image of switch

All of our fabulous servers will communicate with one another through an HP Procurve V1910 Gigabit Layer 2+ smart-managed network switch. The switch has 16 auto-negotiating 10/100/1000 ports and 4 true Gigabit SFP ports.

Comment

Post Header

Published:
2011-01-07 22:33:17 UTC
Translations:
Tags:

Welcome to Revision 3520, up from 3483. This is actually a few different releases bundled together, because we had a little flurry of activity over the holiday period which we didn't post release notes for.

Most Archive users will have noticed that the Archive was experiencing very heavy loads during the holiday period, as Yuletide and other holiday challenges brought high traffic to the site. We know that our sad little 502 page was showing hir face more than any of us would have liked! However, our coding and Systems teams deserve a MASSIVE shoutout for their tireless work keeping the site up and dealing with the performance issues. They worked right through the holiday to keep pushing code improvements and tweaking server setups. We'd especially like to thank sysadmin Sidra for the incredible amount of effort she put into coaxing our poor beleaguered servers into working: her work helped get the number of 502s down from over 9000 (on 25 December) to just 9 a few days later. Go Sidra!

If you're wondering whether the Archive will be this slow NEXT year, then we hope the answer is no! The stress test this year taught us some more about how to configure things for the best possible performance. Even better, we're about to buy some new servers which our hosting service described as 'beasts'. More news on that coming soon...

Highlights

Static pages for collections

As part of our performance enhancements, the wonderful Elz worked till the eleventh hour to put in static pages for collections (you may have seen us mention these a time or three on @AO3_Status). These are pages where the whole page is created once and then saved, so it is served up quickly for the next user. They can't have any dynamic information (comment counts, hit counts), but they are MUCH faster at peak times.

All collections have static pages - the url is http://archiveofourown.org/static/collections/COLLECTIONNAME. They generate when people access the urls, so they are only noticeably faster for heavily visited collections. Check out a few of the following:

Share via Twitter (or not!)

We've noticed lots of our users sharing recs for AO3 works via Twitter. We thought you might appreciate a quicker way of doing this, so we've added a Twitter button which fills in the title and url for you! Just click the 'Share' button to see the button.

We know not all our users will be comfortable with a feature that makes it really easy to share links on external services, so we've included an opt-out - just check out your preferences to select this. If you opt out, it will disable the Twitter option on your works now, and if we introduce sharing options for other external services in the future it will also disable those. (However, opting out won't actually prevent people from sharing links to your work anywhere, it will just make them do more of the work themselves. If you want a higher level of control over who sees your work, you can restrict your works to Archive users only.)

Improvements to kudos

Since we introduced the 'kudos' option a couple of weeks ago, over 51,000 kudos have been left and we've had a ton of great feedback about it. We've fixed a few things that were not quite right with the feature: it's now 'kudos' everywhere (no weird singular usage) and you can opt out of kudos notifications without turning off your comment notifications.

Known Issues

See our Known Issues page.

Release details

Features

  • Added server-relieving caching options to collections and challenges
    • Added static pages for collections (fandom pages, work pages, etc).
    • Added ability to leave kudos from static work pages
    • Ensured that static pages would persist across deploys.
  • Added 'Share on Twitter' button
  • Added the ability to opt-out of allowing sharing of your works on external services (e.g. Twitter).
  • Changed kudos-related wording in notification emails and on work pages to take care of the singular/plural issue - no more strange singular 'kudo'!
  • Removed revision number from page titles
  • More automated tests, continuing our mission to test ALL THE THINGS.

Bug fixes

  • Removed author names from "share" box for anonymous works (whups)
  • Fixed byline bug causing names of anonymous authors to appear in static files and downloads
  • Made it possible to edit a work in a closed challenge without being kicked out of the collection
  • Fixed collection links in anonymous work blurbs showing up as unparsed HTML code
  • Fixed a bug where you couldn't edit tags on a work if they were related to a meta tag
  • tweaked collections pages for display in Internet Explorer
  • Cleaned up story notes section, fixing issues which were causing links to endnotes to be encased in invalid HTML and thus look weird.

Systems changes

  • Added option to turn off kudos notifications
  • Change thinking sphinx delta delayed jobs into an hourly cronjob (relieved load on the server)
  • Changing reindex to every two hours (relieved load on the server)
  • Gave delayed job its own restart
  • Made it possible to run deploy quick on otw2 server (saves some work during repeated deploys)
  • Updating static sweeper

Comment

Post Header

Published:
2011-01-02 22:04:09 UTC
Tags:

Do you love the Archive? Do you have skills in database analysis? Would you like to make these great two tastes taste great together?

We're looking for a MySQL database developer who can help us optimise our database performance on the Archive of Our Own. The Archive is run on MySQL 5.0 and Rails 3, but you don't need to know Rails to help -- we would really like someone who can look over our database query logs and help us spot inefficiencies!

You'd be joining a team of dedicated and passionate volunteers with a love of fandom and tech-geekery. We're looking for someone with the time and skills to help us evaluate and optimise our current setup — this can be a short-term commitment, although if you are able to join us for the long haul your skills will be very welcome.

If you're interested, please contact our Volunteers and Recruitment Committee using 'MySQL database developer' as your subject line, and let us know your skills and experience.

Your Archive needs YOU!

Comment


Pages Navigation