Dear Steve, (Forum getting busy signal)

[Angela]

Originally Posted by seeker5

I have the reverse experience. It's happening less for me. Maybe you're catching the busy signals for me? That's sweet of you .

For you, for you, I do it all for you!

[Savage]

I can see exactly when the spikes are happening because I get emailed a report each time.

There were 2 spikes today, including one about 20 minutes ago, and 4 spikes on May 3rd. As far as I can tell, the frequency of spikes hasn't changed over time.

ServInt is working on this actively, so hopefully they can figure out what's happening. They already ran some diagnostics that revealed no problems thus far.

[ellie]

It's surprising that an host company has not been able to solve the issue, it's been quite a substantial amount of time. It's good to see how patient you are being.

[Brutha]

I have seen evidence that something is going screwy with postgres on the server. That's the only hard lead I have right now. But neither VB nor WordPress use postgres.

How about deinstalling postgres if it isn't used?

[bluelighth]

Is postgres used by anything else? If so, is there any co-relation with that? And would it be logical to try emulating the situation on another machine with the same configuration, especially since it doesn't seem to have much to do with the amount of users at the time?

[Savage]

Good news -- we finally have a diagnosis of what's been causing these load spikes. And fortunately it's a problem that's very fixable.

The problem is caused by a MySQL query in the WordPress plug-in WP-Email. This particular query populates the "Best of..." list in my blog's sidebar, based on how many times each post has been emailed using the WP-Email plugin.

I've had this plug-in running on my site for a while now, so it has gradually built up a sizable database with more than 27,000 entries. That's how many people have used the "Email a friend" form since I last reset the data.

Due to the size of the database, the way this particular query was written, and the inefficiency of the plug-in code, it can bog down the server at random times.

ServInt did some optimizations to the server already, which should help alleviate the symptoms and make load spikes less likely. Meanwhile, I've contacted the plug-in's author, Lester Chan, to let him know of the problem and to suggest some possible code updates that would remedy it. Other high-traffic sites could run into the same problem with this plug-in.

WP-Email is a very popular WordPress plug-in, so hopefully the author will be willing to optimize the code a bit. But if not, I have plenty of other options, such as resetting the plug-in database, deleting part of the database (entries older than a certain date), disabling the sidebar widget portion of the plug-in, replacing the plug-in with a different one, etc.

For now I'm going to let things sit for a day or two to see if there are any more spikes, since it's possible that ServInt's optimizations may be enough. If there are more spikes, I can disable the sidebar widget until I decide on a better solution or get a fix from the plug-in's author.

But either way, we should have this problem solved this week.

Thanks for your patience. I'm sure it was quite a challenge for ServInt to diagnose this, so I'm happy they went the extra mile to figure out what was causing the load spikes. I never would have suspected a WordPress plug-in that has been running fine for years as the culprit.