Dreamhost Outage Highlights Communication Issues

If only outages could be solved with some sort of panic button.

Another day, another outage: hosting provider Dreamhost suffered a loss of connectivity on Monday evening in its Los Angeles datacenter, aggravating customers.

That outage came on the heels of an earlier network transition that also affected Websites hosted on the West Coast servers, and doesn’t appear to have taken that long to resolve: Dreamhost first noted the outage at 5:36 PM PST; by 6:12 PM, it was “confident that the connectivity is back, and the issue is resolved.”

In a bit of irony, Dreamhost had just completed the installation of a backup 10-Gbit Internet feed into its West Coast data centers. “This will further improve redundancy on our primary 10 gigabit lines in the event of failure,” the company said last Thursday.

On Friday, the company began the physical transition from the old connectivity switch to the new hardware, a process undertaken in small chunks to minimize any impacts to customers. Dreamhost moved through the transition without many problems, according to updates it provided at the time.

The problem, according to the site’s customers, was that Dreamhost apparently scheduled the maintenance without first sending a general notice to customers via email, and then noting which customers were affected.

“Rhys Jones,” a commenter on the site, summed up customer complaints:

Our sites have been down for two days now, the fourth separate occasion this month. There is never any notice as to when you plan on taking down our servers other than posting on this page which is obviously an ineffective way of notifying us because it requires us to monitor it 24/7. To make matters worse you don’t even know which sites are on which servers.

You hide behind the fact that your service is cheap and therefore substandard support should be tolerated. The problem with this is that being considerate costs you nothing and people also know that. Most people don’t expect 100% uptime; all they expect is notice of planned downtime so they can work around it.

A)   Work out which sites are on which servers.

B)   Email those clients in advance of any planned downtime. You have

our email addresses so please use them.

That is all you have to do. So simple.

The Dreamhost outage seemed relatively mild compared to the data-center travails suffered by some other companies in recent days. Last week, Microsoft’s Azure service suffered a 2-hour outage in Western Europe, and Twitter’s service also went down. Twitter cited the data center for the issue; Microsoft blamed the outage on a misconfigured network device that disrupted traffic to one cluster.

Earlier this month, dating site What’sYourPrice.com told Computerworld that it had ditched Amazon’s AWS services after a fluke East Coast thunderstorm affected the data center hosting the site’s services. The reason? Lack of communication with Amazon’s support team.

Instead, the site chose a Las Vegas provider, Fiber Hub, which is physically closer and therefore apparently easier to supervise. “We looked at different competitors and alternative servers to host our Websites,” Leroy Velasquez, a spokesman for the site, told Computerworld. “We chose to go with a local one because it’s better for accountability and oversight.”

Dreamhost has a 100 percent uptime guarantee written into its SLA, which will credit users with a day of free hosting for every hour (or fraction thereof) a user’s site is down, up to 10 percent of the pre-paid renewal fee. But in this case, customers on the site seemed interested in complaining about lack of communication. The lesson here? Bringing customers into the loop may not make them happy, but it may make them more accommodating.


Image: mtkang/Shutterstock.com