Gryz

One more update.

I couldn’t believe something was wrong in the google-analytics code. Google isn’t that sloppy. But then I realized that that code does not come from Google. It is part of the “Google Analytics for Wordpres” plugin. And the code is written by a volunteer from the WP community, not by Google.

When I looked at a lot of those websites that had the same problem, I noticed they were all using the “Google Analytics for Wordpres” plugin. So this plugin might very well be the cause of the problems.

I disabled it, and replaced it with another plugin.
Ultimate Google Analytics
Let’s hope this fixes the root of the problem.
We should know in a few days. I’ll update this post.

So there’s three parts to the solution.
1) Use a different GA plugin.
2) Add rules to robots.txt to prevent googlebots crawling for old malformed search-URLs.
3) Wait for old search-URLs to depricate from the google database.

If this turns out to fix the problem, I wonder if I should notify parties involved. Would Google remove the bogus 18 million entries from their database ? Should I contact the author of the GA for WordPress plugin ?

Anyway, Relevanssi had nothing to do with this. It was a very useful tool to warn us that something was wrong. Thanks for all your effort, msaari !

Thanks for the suggestion. I’m an old C-programmer who used to write C-code for networking devices. I have no knowledge about php, and I’m not sure I wanna check out all WP code to see how it hangs together. I was hoping for a log-function of WP, where I can just go through all http-requests. Maybe I’ll see if I can write some code.

I’ve grepped through all the php-code. The only place where I could find the exact string “no-results:” was in the google analytics code.
From googleanalytics.php:

} else if ($wp_query->is_search) {
$pushstr = “‘_trackPageview’,'”.get_bloginfo(‘url’).”/?s=”;
if ($wp_query->found_posts == 0) {
$push[] = $pushstr.”no-results:”.rawurlencode($wp_query->query_vars[‘s’]).”&cat=no-results'”;
} else

It looks like the string “no-results:” is pre-pended to the search-string. This seems like a place where excessive no-results: could be prepended.

I disabled google analytics for a few minutes on our website, and I still saw new searches with the mangled query. 🙁 It’s very weird. I’ll look into it again this weekend.

The problem is happening at many sites.
When searching on google for “no-results:no-results:” I’m getting 15.8 millions results ! Although google only gave me 355 results. Still doesn’t look good. I’m surprised nobody ever looked at this before.

http://www.google.nl/search?complete=0&q=%22no-results%3Ano-results%3A%22

Is there a way to see where searches are coming from (IP address or domainname). When I reset the Relevanssi logs, I’m getting log-entries with no-results: in the query-string within minutes. I can’t believe it’s google’s webcrawler that is so quick to crawl my website. However, if it’s not google, then how does google pick up those bogus URLs ?

Also, when I disable Relevanssi, is there a way for me to see the query-strings that get processed by the default WP-search engine ? That would allow me to prove to myself that the bogus searches also happen when Relevanssi is disabled.

We are using the latest non-beta version (as usual). 3.2.4.

It seems XML sitemap generator uses a spike of a few MB while creating the sitemap. That’s fine, I guess. XML sitemap generator was the plugin that showed me there is a problem with memory on my server. But that doesn’t mean it is the cause.

Once I realized that, I could remove my post here. Or keep it, and hope that someone would read it, and make a useful comment. I’ve spoken with a friend (who is not very technical, but does have a lot of experience with building websites). He told me that my webspace-provider might have installed a lot of stuff in apache or php, which can be statically linked to the executables. And thus cause a growth of my webserver process. Without me having any influence on it. I think it’s weird. Because that means they sell webprocess with 48M ram, but already use 25-30mb just because of apache/php.

At the moment the website runs fine. I disabled 1-2 specific plugins, and the normal memory usage is 42/48M. I am now waiting to see if the server ever restarts/reboots. Then I can maybe see if we have a memory leak, or whether the high usage is normal.

I have a little more data.
I installed the TPC Memory Usage plugin.
It seems the website is using 40-45MB constantly. So the number reported by this XML sitemap plugin is not memory usage of the plugin only, but usage of the whole process.

Still, adding 2500 tags to the sitemap adds 3-5MB to memory usage. (Or more) And that pushes memory usage over the 48MB limit. Not good.

Now I am wondering: is a base memory usage of 45 MB is normal or not ? Too bad the TPC Memory Usage plugin does not specify how much memory each plugin uses. I guess I’m gonna disable plugins one-by-one to see what impact on memory usage they have. But I can only do that during the “quiet hours”.