We have been hosting BGT at CWI hosting for at least three years now, and they have finally succeeded in running us off.
For much of that time we were happy with their service, they would reply promptly whenever we would submit a trouble ticket. Their prices remained high for the amount of bandwith and disk space we were allocated, but as long as we were getting good service there wasn't much point in switching just to save a few bucks. Not to mention, we were still well within our bandwidth limits, so why tempt fate by moving to a different server.
It was really tempting to call for a move last year when they tightened up the security on their server, which made it more difficult to manage some aspects of the site, but we stuck with them.
Then, right around Thanksgiving, with no warning at all, they shut down our site!
Not only were we not given any warning, we were not contacted after the fact either! We had to call them!
When we did get in touch with them, their reply at first was that we were using too many resources, such as disk space. In fact we did run out of disk space the prior week, but I cleaned it up and we were back down below half our quota.
Then he mentioned bandwidth and possible processor overuse, but there were problems with that claim. We have never had a single day where our bandwidth has exceeded 1/30 of our monthly quota. in fact, we run about 2/3 of that level. As for processor use, in November we were serving up only about 80% as many pages as we were serving up during the summer. Our usage was down, and our pages were serving up fast.
Anyway, in the process of taking down the site, they also broke it. The replaced our .htaccess with their own, and deleted ours without making a backup. They also broke our mailing lists, and claimed that they were woking fine, and turned off our shell access.
CWI then sent us this: "We have found that your account backpackgeartest.org has a high disk usage exceeding a few gigabytes. Although your account allows this and more, we will be moving your account to a server that can accomodate your sites growing needs. This transfer will be taking place within the next 72 hours."
Okay, it would be a pain in the butt, but we would suspend uploads to the site for three days. We would be on a new server and we would all be happy again.
Notice that they said it was the disk usage that was at issue, that the new server would accomodate our "growing needs" and that it would happen within 72 hours.
Well after nothing happened for more than a week, we pestered them about it since we wanted to start uploads again. It seems that it was our fault that they could not move the site "Due to the extent of the directory structure of your account, archiving and transferring your account causes load spikes, increasing the transfer time and decreasing server stability."
Oh No!!! Our oh so complex directory structure! That tough directory structure makes about a 250 MB .tgz file in an account that is supposed to have 3 GB available. If making the backup is causing a load spike, I would suggest that they hire some techs that know what they are doing. All UNIX and linux systems have a program called "nice" which allows you to run something like a backup without putting as much of a load on the server. Directory structure does not make a damn bit of difference when it comes to making a backup.
So, when they finally get it all done and transfered again, they have once again broken the site! Their fancy backup program breaks the symlinks. They can't just to a tar, they have to do a tar that tries to outsmart the system, and breaks it in the process. Having had enough of their incompetence, I went ahead and fixed all the symlinks on my own.
Then came what I consider to be the real reason for this mess, they wanted to sell us a dedicated server, which definitely isn't in the budget, and if it was, it would not be through a company that handles issues the way that CWI does. Here are some choice quotes:
"This domain has been moved to a temporary staging server recently as it has been causing a high load on the server which it is hosted as well as using a large amount of resources(CPU/RAM)."
Fine, but why does the story keep changing? Why can't you deal with us nicely and try to work out the problems?
"It seems that there are several factors contributing to this problem which are that your sites JSP code tends to jun at high amounts of RAM and CPU which require it to be restarted regularly through the day,"
Well, that sure seems serious! Uh, except for the little fact that we do not run Java Server Pages (JSP). Oops!
We run PHP for all our pages. While it is possible that we are using a lot of resources with PHP, this whole line of reasoning is called into question when you look out on the web at some of the other complaints. It seems that this is a standard claim that they make when they are trying to sell dedicated servers.
Oh yeah, I ran a load test on my old 266 MHz PII system running linux, PHP and MySQL and the load put on that system by the number of hits that we get is below 5% for the day. I am assuming that they are running servers with a little more horsepower than my ancient box.
"large amounts of email that are sent from the sites mailing lists,"
This is the only part that seems to have even the slightest ring of truth to it. But I even tend to discount this for two reasons. The first is that they have not been honest or consistent with their other reasons. The second is that they never mentioned it before, or did anything to see if we could work out the problem.
"and finally directory structure and space."
Oh, that whole directory structure and space thing again. The directory and file system is the fasted database you can have on a computer. It may sound good to people that don't know any better, but they are full of shit on this one.
It seems that when they sell you a 3 GB account, that it means that you better not go above 1.4 GB or they will shut you down. Did I mention that a large part of that 1.4 GB is because they do not have the mailing lists set to automatically clean up attachments that are sent with messages? They just build up in a folder that applies towards your account, but you have no way of cleaning out. The way I found out about this is that I lookd at the TAR created by the backup program. It had crap in there from years ago.
"It seems that your site has grown beyond the normal scope of a shared environment and is now in need of its own dedicated environment to allow it to continue to grow and function. We would recommend moving to a dedicated server very soon."
There is the crux of it. We want you to give us $200/month instead of $30/month.
So, we decided to give dreamhost a try. Upon reading their TOS and reviews, it appears that they actually try and work with you when your site is generating too much of a load on the servers. They even have a system that will automatically send you warning emails when you are within 10% of exceeding your storage, bandwidth or CPU usage, and they give you a month to work things out before pulling the plug on your site. Somehow I trust an automated reporting system more than I trust ever-changing reasons given by someone trying to get an extra couple hundred dollars a month.
In the process of changing things over, things got really slow on the server at CWI. We could not eve get the control panel to respond. When we complained we got this reply:
"This slowness you are seeing is due to the fact that your site is still causing a high load on the temporary server. Therefore the load is much higher than normal and services like FTP, Cpanel and Email are slow or delayed."
So, once again, it is our fault, case closed.
So I turned off our site. I put a "deny from all" in our .htaccess file. Anyone that went to our site was immediately given a 403 access denied. We weren't producing much of a load on the mailing lists either because they were all basically down, as was our email.
Of course, you would expect that would solve the problem, since it was our site causing the problem, but alas, it seems that people just accessing our site and getting just headers as a response was still too much for the system.
Yup, serving up 4000 headers a day is enough to bog down their wonderful servers. And yes, they did claim that the traffic was still going to the site, so that was the problem.
But low and behold, after we had changed the DNS, and I was trying to get the last few pieces of the site moved over, I was told on the phone that they could not run a command on that server for me because another tech was working on it. It would be run later. How could that be that someone would need to be working on that system when all the problems were caused by our site that was no longer running there? Hmmmm?
Then the next day, while I was logged in to that server through ssh, a funny thing popped up, a message that they were upgrading cpanel on that box. After that upgrade, cpanel seemed to run fine, and up to speed. What do you make of that?
You also might want to read the review of what happened to another site at CWI. It might sound very familiar. Read more about it at epinions.
Update 02-02-2006: We have now been running at Dreamhost for a month. They have a system set up so that if you use more than 50 CPU minutes in a day, they will send you an email, letting you know that you are coming close to exceeding your fair share of the processor. We have yet to receive any such warnings.
So if we aren't getting any warnings from dreamhost, or traffic is at the same level, that just lends support to my belief that CWI was making things up in an attempt to sell us a dedicated server.
Oh yeah, Dreamhost costs 1/3 as much as CWI.