The Trouble with Salsa: Inside Yesterday's Dem Software Platform Woes
BY Nancy Scola | Thursday, May 20 2010
Salsa Labs is, today, reconsidering having one line of connection through the Internet between its software-as-a-service platform and the many progressive and Democratic groups that uses its messaging and organizing software. At least one client, though, is wondering if that examination is overdue. "We left a lot of money on the table yesterday," said a representative from one Democratic organization, a Salsa client.
The trouble is that yesterday, the day after some fairly major primary races in the U.S., Salsa Lab's eponymous flagship software project experienced a complete outage lasting at least five hours that left its many clients totally unable to use the platform. Salsa's client list, a sort of who's who of progressive politics, includes the AFL-CIO, ActBlue, Chuck Schumer's re-election campaign, the DCCC, and many others.
Justin Nemmers, Salsa Labs' Chief Operating Officer took responsibility for the problem, saying, "We should have done better, we could have done better, we will have done better."
The nuances of exactly what went wrong with Salsa yesterday are, frankly, complicated, and a little cloudy at this point. Your faithful blogger spent a few hours today going down the rabbit hole of multiply redundant Internet service providers, fiber optic back haul, Border Gateway Protocol and other, well, excruciating topics.
But here's what we know
Salsa's Nemmers, while, again, acknowledging that Salsa Labs let its customers down yesterday, traced the problem to the company's London-by-way-of-LA Internet service provider, Packet Exchange.
"The situation," said Nemmers, who previously managed government services at the open-source software company Red Hat, "was that in working with our ISP, we were under the impression, we had been told that we were multiply redundant," or, in other words, that Salsa Labs were paying their ISP for more than one path to the rest of the Internet. Should one footpath from Salsa to the rest of the world get washed out, the thinking goes, the other paths should remain open and viable. "We've never had a reason to believe otherwise," said Nemmers, about his expectation of redudancy. "We've never had an interruption like this. We've taken all the best practices and reasonably accepted steps. But a bunch of things came together and ended up causing us to go down."
That 'bunch of things,' says a PR rep for Packet Exchange, Salsa's ISP, traces back to a FiberLight, the Georgia-based fiber optic cable company that Packet Exchange relies upon for its Internet service in the Washington DC area. According to the rep, a construction accident yesterday at a Reston, Virginia data center ended up with a cut FiberLight cable, which cascaded down until Salsa Labs found itself without service. "Outages happen," said the rep. "Fibers get cut." Digging, for example, can often rip up existing cables. One of the worries floating around the cyber security world is that the deliberate targeting of the cables that together form the Internet could leave Americans disconnected as Salsa customers were yesterday. A rep for FiberLight did not return a request for comment left earlier today.
Packet Exchange's rep was not able to respond immediately to questions about Salsa's expectations of multiply redundant service from the company. Technicians at both Fiber Light and Packet Exchange, she said, are still digging through yesterday's events to figure out what went wrong.
Nemmers, for his part, says that Salsa Labs isn't taking chances from here on out. By the end of the day today, or tomorrow morning at the latest, said the COO, Salsa Labs will have up and running two completely discrete connections to the Internet. Internet routing work, including the use of the Border Gateway Protocol mentioned above, will blend the two systems together. "We'll be in a situation where if there was a second interruption there would be no interruption to our customers," said Nemmers. "We're working until it's done, and we're not stopping until it's done."
Because Salsa's own email system was connected to the problematic ISP, Salsa Labs wasn't able to reach its customers by email yesterday to discuss the troubles. The company found another path. Said April Pedersen, Salsa Labs co-founder, reached via email today, "We directed all of our pages to our Twitter stream and provided real-time updates for the duration of the outage."
Related posts: "Spicy Industry News: AFL-CIO Taps Salsa"