UESPWiki:Upgrade History/2008

This is an archive of past UESPWiki:Upgrade History discussions. Do not edit the contents of this page, except for maintenance such as updating links.

19 December 2008

• Removed the AdBrite ad JavaScript code reverting soley to Google Adsense.

11 November 2008

• Reenabled mod_gzip on content1.

9 November 2008

• Checked drive space on content1/db1/squid1. db1/squid1 running under 50% usage (nothing done). Content1 was running at 74% drive usage before some old logs were deleted. Currently running at X%.
• Fixed conflicting server-id between content1 and db1 which was preventing replication from occuring (likely the cause of some of the missing forum posts when db1 was switched on). Mirroring of the UESP wiki confirmed and should finally be able to switch it to use db1 in the near future.
• Cache utilization on squid1 is still around 65%.

21 August 2008

• Removed blank line at bottom of ParserExtensions.php which was causing an RSS validation error.

27 March 2008

• Setup a redirect on content1 from www.uesp.net/phpbb to forums.uesp.net to try to prevent caching issues of old links to the forum.

21 February 2008

• Set the $configdate in LocalSettings.php on content1 to force invalidation of entire Wiki cache. This will force the various changes done over the past few weeks to be applied to everyone.
• While Google analytics was removed from the site the cookie will still be present in returning visitors browsers which will prevent the main content page from being cached for anonymous users. Looking at Squid settings there seems to be no simple method to ignore the presence of a particular cookie.

20 February 2008

• Installed and enabled Tidy on the Wiki.
• Installed the ProtectSection extension as requested.
• Disabled the use of Google Analytics as it may have been interfering with the caching of the main content pages due to use of cookies. Changes will likely take a while to propagate through the cache but will keep an eye on the cache utilization (currently averaging 60% cache hits) and see if there's any improvement.

19 February 2008

• Reduced MaxRequests to 1000 from 10000 on content1. This will cause the Apache child threads to restart more often reducing their memory consumption some.
• Removed "ignore-private" on squid1 for default objects. This might/should eliminate issues with logging in that have been reported.

17 February 2008

• Turned off KeepAlive and compression off content1 to try and get some better performance out of it.
• memcached on content1 is up to 120MB of memory.
• Removed sidebar ad unit on wiki. It has lower click thru than bottom ad unit and removing it cleans up the display, especially for short pages.
• Removed Google search on sidebar (barely used).
• Removed top text ad unit on forms and added bottom image ad unit.
• Added "forums.uesp.net" option in Google search box on the forums. Fixed issue with search form not working on topic pages.
• Added the scripts countconn and countips to squid1 for easily checking the number of incoming connections.
• Turned on Apache access logs on content1 with a daily rotation for more accurate traffic count.
• Tweaked some MySQL settings on content1. Set thread_cache from 0 to 8, thread_concurrency from 2 to 8.
• Reverted the CSS settings in the forums for the color of the main scroll bar (from beige to whatever the default is).

7 February 2008

• Optimized 3 images on the main page (2 logos and background) by saving as JPEG. Went from 96.8 kb to 18.0 kb saving 88.8 kb. This reduces the total size of the main page by almost one half (from 181 to 92 kb)!
• Added expires headers on content1 to some files that didn't have it (/w/images and /w/index.php).
• Tried again to remove the ETag header from all files. Some files still have it for despite it being completely disabled in Apache. Perhaps MediaWiki is adding the header or it is a caching issue( not a big deal anyways).
• Minified wikibits.js and ajax.js with JSMin for an overall savings of 14kb (5kb after compression).
• Testing site in YSlow yields a score of C(74) compared to a few months ago we scored a solid F(?). The way things are scored we are probably closer to a B (Google ads, Etags, CDN all the lower score) and with a little further tweaking an 'A' might even be possible.
• Note that moving wikibits.js to load from the bottom of the page rather than the top resulted in some minor display issues in Firefox (change was undone).
• Set $wgSquidMaxage to 2678400 (31 days) from the default 18000 seconds (5 hours) on content1. Could not figure out how to set it to 0 like Wikipedia.

6 February 2008

• Installed mod_gzip on content1. All text/html, js, and css pages should be compressed if the client supports it.
• Turned ETags off on content1.

5 February 2008

• Turned KeepAlive on for content1 (1 second timeout).

4 February 2008

• Re-enabled the daily auto-restart of Apache on content1 to clear any lingering connections.
• Reduced MaxClients to 60 on content1.
• Enabled memcached on content1 with an intial size of 256MB. Will monitor performance and adjust size as needed. After 1 day memcached memory use is up to 50MB.

1 February 2008

• obmap.uesp.net and simap.uesp.net forward to the relevant map (shorter URLs to remember for people wanting to quickly get to the maps). maps.uesp.net currently forwards to the Oblivion map but may eventually go to a general map/index page leading to all available maps.

29 January 2008

• Edited various Apache ExpiresDefault and Squid refresh_pattern settings in order to cache more cacheable content. May need to still make changes in order to make sure any cacheable stuff is properly cached and any uncacheable stuff is not. Short term saw a small increase in cached objects from 50 to 65%.

28 January 2008

• Re-enabled ob_gzhandler in LocalSettings.php.
• Changed namespace spelling from "Talk" to "talk".
• Random pages should no longer display talk or any pages with a "/" in them.

25 January 2008

• Fixed issue with Oblibion/SI maps not showing images.
• Installed sar (SysStat) on both content1 and squid1 servers.

24 January 2008

• Switched the DNS entries for uesp.net and www.uesp.net to point to squid1.uesp.net. Squid settings will still have to be checked and confirm its caching things correctly. As it is now it might be just passing most requests through.
• Disabled the mod_limitipconn and mod_evasive Apache modules as they were causing issues with the Squid cache. Will need to confirm and check settings and renable at some point.
• Reset the Apache Timeout setting to 300.

23 January 2008

• Enabled mod_expires in the Apache web server and made some initial settings as listed on the talk page.
• Turned on the Squid settings in MediaWiki.
• Received the squid1.uesp.net server and setup the Squid cache. While it is currently working it seems it is not caching the Wiki pages (requests go straight through). Note that uesp.net still points to the main content server (content1.uesp.net) and won't be switched to squid1 until it is confirmed as working.
• Changed Timeout from 300 to 60 on the Apache web server.

15 January 2008

• Changed wait_timeout from the default 28800 to 90 seconds in MySQL.
• Changed the ad panel to a single 120x600 unit (text and image).
• Added ad panel at bottom of page (image ads).
• Turned KeepAlive off.

14 January 2008

• Ran Optimize Table on all tables in the wiki database. Tables that seemed most in need of it, based on time required to run optimization, were searchindex (34.5 seconds), pagelinks (13.2), and revision (12.8).

9 January 2008

• Gave the admin group write access to Oblivion's alchemy, map, and simap directories.

8 January 2008

• Enabled KeepAlive on the Apache web server with a timeout of 1 second to observe any change in performance.

4 January 2008

• Tweaked web server settings. Increased MaxClients from 75 to 100. Changed MaxRequests from 0 (unlimited) to 10000.
• Disabled gzip compression of Wiki content (unsure if this has any effect with current web server settings).

3 January 2008

• Reset $wgAjaxSearch = false
• Reset $wgDisableCounters = false (no performance change noted)
• Reboot server for the first time in 6 months to see if performance issues are resolved.
• Disabled web server automatic restarting every day (testing performance after reboot).
• Setup ssh/sudo access for Nephele to be able to restart web server if needed.
• Apparently the mail setup was allowing anyone to send mails through the server. Temporary shutdown smtp to note any performance affects (needs to be permanently fixed).

Prev: 2007

Up: Upgrade History

Next: 2009