Hosting company tech support

[Monkeyleg]

It's not easy finding a good company for hosting websites. There's a lot of fly-by-night operations out there.

In 2008 I finally found a good company. 24 hour a day live tech support, 99% uptime...things were good.

Last September I started getting an unusually high number of 500 errors. These are the errors you get when there's a problem with the database or the server.

The number of errors went from a couple dozen a day to hundreds or, in a couple of cases, a couple of thousand a day. Not only was this a problem for my visitors, but it also had the potential to severely damage my rankings on Google. (My fears were borne out when my Google traffic dropped 8-13% in November and still hasn't recovered).

This was also where the usually reliable tech support people became a real PITA. They couldn't figure out what the problem was, denied there was a problem, said it was my fault, said it was hackers, or just told me to deal with it. I finally had to call a sales person and threaten to take my site elsewhere before they assigned one techie to find the problem.

Finally, it was fixed. I was back to just a few 500 errors a day, sometimes even just two or three.

On Friday my stats program was on the fritz, so they had to reprocess log files to get it working. That took about a day and half. I got an email this morning telling me the stats were working again. So I checked it out. It showed 812 500 errors for yesterday, and 93 errors as of 8 this morning.

I called tech support, and the guy told me it was probably a coincidence. I told him the long story about the problems in September and October, and he told me it was probably a coincidence that the number of 500 errors went from hundreds down to dozens or less.

I hate being treated like an idiot, and this guy was giving me the red carpet idiot treatment. If an electrician does some work on your house, leaves, and there are sparks flying from the outlet behind the TV, chances are it's not coincidence.

I did some searching in the log files, then compared the timing of the errors to the Event Viewer on the server. The reprocessing of the log files was using too much CPU and memory, causing spikes where visitors couldn't get pages to load.

Where do they get these guys?

[Headless Thompson Gunner]

Good geeks are hard to find.

[mtnbkr]

There's a lot of fly-by-night operations out there.

Yes, there are.  I interviewed with one last year (Director of Operations) and never pursued things after that initial interview because I didn't expect this company to be around for more than another year or so.  Shame though, the work could have been interesting.

In 2008 I finally found a good company. 24 hour a day live tech support, 99% uptime...things were good.

I wouldn't call 99% uptime good.  That's more than 3 days of downtime a year.  Uptime should be in the 99.9%-99.99% range.  I created and ran a managed service (not web) for years and achieved 99.999% without undue strain (required a hot failover system, but the service was never completely unavailable).  I would find a new hosting company.  This is your business after all.  

The reprocessing of the log files was using too much CPU and memory, causing spikes where visitors couldn't get pages to load.

They need to be logging to an external system so the http server isn't doing anything with logfiles other than sending them elsewhere.  Then, you could run your own reports against the log files without impacting the production box.  

Chris

[CNYCacher]

If you ever do want to take your site elsewhere, look me up.

[Monkeyleg]

I should have used numbers after the decimal point. Uptime is usually 99.9%. There are problem times every few weeks where the uptime drops to 99.7% or 99.8%

They need to be logging to an external system so the http server isn't doing anything with logfiles other than sending them elsewhere.  Then, you could run your own reports against the log files without impacting the production box. 

Makes sense to me. I'll ask why they don't do that.