Our Hosting Problems

by Richardson ~ May 27th, 2006. Filed under: Administration, Blogs & Blogging.

As many of you no doubt noticed, The Korea Liberator was down for over a day. This is going to explain what happened and why, with some WordPress jargon, some technical detail, and griping (ok, ranting) about support, so if you aren’t interested, no need to continue.

Update: Some of the deletions and changes have caused the ‘Post Comment’ button to be inoperative (thanks for the heads up, lalpuh)… I’m working on this. It may come and go, and I’ll have to sleep at some point, but it will work in the end!

Original post: We had an issue last week when our service provider, LunarPages, suspended our account for several hours after our CPU usage apparently exceeded the 1 percent threshold (that is not publicized in the service agreement and cannot be checked by users).

At the time I thought it was a WordPress contact plug-in I was testing, which I deleted. Support unsuspended our site and sent this, “Everything seems to be resolved as of now according to your account information…” on 21 May.

On 26 May we received our first warning in the middle of the night, followed by suspension two hours later. We were told that our usage from 21-25 May far exceeded the threshold, and we were crashing their server with PHP processes. Which ones, they could not tell us:

Please note that the above mentioned resource
usage is due to PHP processing. Since we do not have the exact
scripts to provide (since the manner in which PHP scripts are
provided on the server only shows the username and not the script
name causing the high usage), you would need to review your
account for scripts that may be the cause. Please check ones such
as forums, blogs, content management systems, and galleries as
these are especially likely to create high usage.

Basically, ‘one of the needles in your haystack is out of alignment – we can’t tell which one, or even the proper alignment, but you need to fix this problem, thanks.’ Not the most helpful you’re-suspended message one could hope for, eh?

So even though on 21 May we were told things, “seems to be resolved,” in reality the entire time things were not ok, and we had two hours warning in the middle of the night before being suspended. Even one day prior notice and I would have been able, I think, to resolve the issue.

As we have been receiving several hundred to several thousand spam comments and/or trackbacks per day, which Akismet plug-in is catching very well, after a little research I thought that the overload might have something to do with requests by spam-bots.

It also seemed that many spam were utilizing the comment preview capability to do this – so I promptly removed that avenue. I was able to determine that with log information provided by the single useful reply from a support technician (greatly appreciated), which also pointed me to caching plug-ins as a possible solution.

At any rate, with the account suspended we only have cPanel access, which allows the deletion of suspected offending files, and the upload of plug-ins that might help save the day.

However, without the site up and active, it is impossible to actually activate those plug-ins – meant to help reduce spam/reduce server load – and test to see if it might work.

With a couple hours of the site being down, I have found, customized, and uploaded the following pluig-ins:

Bad Behavior 1.2.4 (anti-spambot)

HashCash 3.1 (anti-spambot)

WP Cache 2.0 (reduces PHP processes by caching pages)

But I could not activate them until LunarPages support unsuspended the account. (Technically, I could have, since they gave cPanel access; they ‘suspend’ by changing the .htaccess file, which I could have simply changed back. However, that would have likely earned an end of service, which would mean locating a new host and days of downtime while moving.)

I sent multiple emails and posts (even a new ticket) to support explaining that a) I’d deleted the suspected offending files via cPanel; b) uploaded new plug-ins recommended by the support tech who understood the problem (and was then apparently off this weekend), and; c) needed the site back up in order to activate/test the plug-ins. I got back replies like this few times:

Cpanel access has been granted. I am able to login at…
Please let us know when you have completed the fixes. We will
monitor the usage to see if it helps and hopefully get you back on
a production server once it comes down.

That was late yesterday, then no answers until today. Similar from three support “technicians.” Obviously they did not really bother to read what I’d written, which means the site was down for the greater part of a day because their techs failed to read and comprehend the basic outline I’d sent them.

Finally I got some attention with a simple, one line question, “As I have cPanel access, do you mind if I alter .htaccess?” (I didn’t know I’d gotten any attention for a few hours after the fact, as my hotel here in Barcelona had a power outage, and of course the net was down – the Spanish just don’t take the internet as seriously as Koreans).

So someone finally reactivated the site, and I activated the plug-ins. Spam-bots did get past WP Cache to Akismet (which caught them – about 700 in two hours), but when HashCash was up and running (correctly), the spam-bots seem to have been stopped even before Akismet. I hope.

I’ve been with LunarPages (via DPRK Studies) since 2001 with excellent support the few times it was needed, but they really dropped the ball with this one by simply not having techs that a) took the time to read what the customer (me) wrote to them, or b) the technical inability to comprehend what was being said, even if in plain and simple language. Or maybe a combination. It could have been worse, or even much worse.

Now, I do understand that if our site is crashing the server, it has to be taken down. Got it, that’s completely fair. BUT, not warning us sooner, and then seeming ignoring us for the better part of a day is undeniable negligent. That, along with incompetent support (most it seems, but not all of the techs) are the issues.

The bottom line is that if you have a simple website with mostly static html used for relatively unimportant tasks (family, email domain, etc.) LunarPages is great; if you’re going to run a blog you may want to look elsewhere. I think we’ll give them another chance since it would be a huge pain to move (including the domain) someplace else.

If this does happened again, you can contact us via the DPRK Studies site, which has thankfully been spared by the spam demons up to this point, and thus seems not to have violated the unwritten but apparently sacred 1 percent of CPU usage commandment.

Also, if you know of any hosts that can actually handle a WordPress blog with a fair amount of traffic, please comment about it.
—–

2 Responses to Our Hosting Problems

  1. Richardson

    test

  2. Richardson

    another test…

Leave a Reply

Subscribe without commenting