[wp-hackers] Running a network

Carl Roett carlroett at gmail.com
Sun Dec 30 22:10:11 UTC 2012


Given sufficient resources, WordPress multisite is usually well-behaved on
large networks.

Typical causes of multisite performance problems:

1) Hundreds of thousands of spam profiles on a site that are flagged as
spammers. Each of those spam profiles consumes ~10 DB tables. We had one
site with 100K+ auto-flagged spam profiles on it that had over *a million*
db tables. Deleting those tables dramatically improved site load times.

2) Excessive crawler hits, causing memory starvation on the server, aborted
PHP threads,   SQL server throttling, and bandwidth throttling. In one case
we had Google hitting a large site dozens of times a second, causing the
active PHP threads to consume the entire 1GB cap on the server. The site
owner accidentally triggered the crawler flood by changing the templates on
their site, which changed the page structure on their site, and caused
Google to re-crawl 50,000+ pages. To somebody that can't view the server
stats, the site would load incredibly slowly and possibly throw error pages
from time to time.

3) Javascript code that operates on multisite profiles but not the root
blog, and which blocks the DOM ready event until complete. Many possible
sources including plugins and themes.

4) However, if:

a) You had multisite up and running and could view the individual user blogs
b) Then you enabled wildcard domains
c) Then the site became very slow

That means there's something wrong with your DNS settings. Something is
timing out, then hitting the next node up and finding what its looking for.

^C^


More information about the wp-hackers mailing list