Uh oh. All of your chillers have just shut down at once. How much time do you have to diagnose and fix the problem before the data center heats up and equipment begins to fail? For most data center managers, the answer is "not nearly enough." That was the case for the staff at Hosting365, one of Ireland's largest hosting companies, when all seven of their chillers shut down at 1 pm yesterday afternoon.
Hosting365's main data center is at capacity (they're currently expanding it) and packed with customer equipment. Almost immediately the temperature began to rise by about 3.5 degrees (2 degrees C) per minute. Within 15 minutes areas of the data center were experiencing heat above 40 degrees Celsius - just over 100 degrees Fahrenheit. Servers began to shut down, and staff turned off the rest to protect the equipment.
Hosting365 had figured out the problem - an electrical short in a fan coil, which then fried a fuse that supported the other chillers - within 10 minutes of the original failure. Within 20 minutes, Hosting365 staff had replaced the fuses and brought the chillers back online. By then it was already too late.
"It's clear from this issue that the suite cannot tolerate even an 18 minute failure of the chillers," Hosting365 Managing Director Steven McCarron wrote in a status update for customers. "We have an excellent electrical and facilities team and we'll be looking at ways to beef up the cooling capacity and redundancy. Ironically, we've spent more money than the company makes in a year on improving and adding redundancy to our infrastructure in the last two years, and then, as is often the case, something small springs up and causes a problem."
To its credit, Hosting365 acted quickly and provided customers with a prompt review of the incident. "As managing director, founder and owner, the buck stops here," McCarron wrote. "So, let me detail exactly what happened today, the order of events, how we reacted and handled the problems and what we're going to do about it tomorrow."