Wikipedia:Wikipedia Signpost/2005-06-06/Global downtime
Downtime scheduled Tuesday for server relocation
Wikipedia and all other Wikimedia sites will be unavailable for a period of time on Tuesday, 7 June, as the main cluster of servers is moved to a new location. This follows a brief period last week when the sites had to be put into read-only mode.
Last Saturday, Wikimedia Foundation President Jimmy Wales announced that at 7:00 a.m. (UTC) this coming Tuesday, the servers located at the Neutelligent colocation center in Florida would be moved to a new facility across the street. The exact length of time for which Wikipedia will be unavailable is not certain, but Wales said "the site will absolutely be down for awhile." Employees of the colocation facility will handle the move with assistance from Wales, along with a few other Wikimedia volunteers and staff.
Mark Williamson wondered why the servers were not being moved in two or more shifts, so that one group could always remain available. In response, Wales explained that "It was a difficult decision", but the complexity of such an operation, and the risk that mistakes could produce an even longer period of downtime, favored handling the move all at once. Site performance would also likely suffer from extreme slowness with only a partial set of servers.
Meanwhile, the Foundation has added its third data center with the addition of eleven servers in Amsterdam, hosted by Kennisnet. This joins a group of servers installed last December in Aubervilliers, France to provide a squid cache. These servers may be used to host a site message during the downtime; however, because the main database servers are located in Florida, the Wikipedia database will still have to be offline.
For Wikipedia editors (as opposed to readers), it will be the second period of unavailability within a week. After suffering performance problems on Monday and Tuesday, Wikipedia was set to read-only mode for about two hours last Wednesday for server maintenance. During the maintenance, security measures for user passwords were also upgraded (see related story). Reports of site performance issues virtually disappeared after the maintenance was completed.
Discuss this story
"Neutelligent has huge fiber commitments with the following carriers, totalling over 4.5 gigs: UUNET, Level 3, and EPiK. By doing so, that will ensure 0:00 down time, not 1%, we mean 0% downtime." [1]
Oops, I guess that 0% is rounded or something. Smyjpmu 23:52, 4 Jun 2005 (UTC)