 | Site issues updateFebruary 1st, 2007 |
As many of you have noticed, the site is periodically (as in, multiple times a day!) becoming unavailable with an “account suspended” message. So here’s the status:
A while back, apparently, my host put a tighter CPU limit on me because the site was getting so popular it was affecting other sites on the same server. You would not be alone in thinking this an odd way to handle the situation: “Hmm, this site is pushing at the limits of its allotment. So let’s decrease it.” They of course neglected to tell me that they did this. In fact, they neglected to even note it in my account, so all subsequent support folks have been baffled. Kudos to support guy Jamie for working to figure out the problem. Unfortunately, it’s still not fixed.
There’s now a support ticket in asking for the normal CPU limit to be reinstated, plus a request to move to a high capacity server (which is a hosting plan they offer). But they may not have a higher capacity slot available right now. In which case, we’ll be moving to new hosting anyhow. I’m waiting for them to get back to me — they are now a day overdue on the answer, so I will be calling them tomorrow. Many thanks to all you readers who offered hosting advice or space.
Lots of folks were wondering what exactly the traffic and site looks like, so here’s an infodump:
- The issue is CPU load caused by WordPress SQL queries. Not fancy ones, either — the standard ones.
- January cumulative stats were 84000 sites, 240000 visits, 540000 pages served, 2.2m files, and 2.7m hits.
- January average daily stats were 86000 hits a day, 70000 files, 17000 pages, and 7700 visits a day.
- It’s worth pointing out that at the start of January we were at around 6-7000 visits a day, and 65-90k hits. At the end of January, it’s more like 8-10000 visits a day, and the hits were trending in the 80k-160k range.
- January was the busiest month ever, and represented a big increase in sites and visits over December (21% more hits, 46% more sites, 19% more visits)
- It probably would have been more had the last 1/4 of the month not been stunted by the site issues.
- Bandwidth usage in January was 43 gig — I have a 100gig allotment per month with my provider.
- The blog database is around 4 meg all told, which isn’t too bad at all. There’s 13,000 records in the comments table though!
- That said, the rest of the site including all the presentations, artwork, music, blah blah blah is around a gig.
This being the only site I’ve ever run other than the book’s website, I have no good context for these stats. 
So I don’t know whether the blog traffic is indeed so high that it’s to be expected that I would be maxing out my shared server situation, or whether something seems clearly misconfigured somewhere.
On a related note — I did install WP-Cache2, but I notice that when I check what files it has cached, it only ever seem to have the pages ending in /feed cached, or the front page. All the other pages (like, every post!) do not seem to be caching. I have no idea why, and would love an answer since that should help the load enormously. (It would also help if the rest of the site, which gets hit quite a lot, were in the WordPress CMS, but alas that would break everyone’s links and it’s a huge task to move everything over!).
One positive note: a bunch of the people reading on feeds that seemed to drop off a few months back seem to be back. I assume that somehow I had broken the stats tracking through FeedBurner, because post-upgrade to 2.1, the numberleapt back up by several hundred.
Lastly, I think we will be redoing the visuals and trying to reduce clutter soon. Anyone who’s got ideas, go ahead and submit ‘em. 
You can follow any responses to this entry through the RSS 2.0 feed. Responses are currently closed, but you can trackback from your own site.
