Google Cloud freaked developers out when it threatened to automatically and permanently lock up a company's critical app
- A Medium post got a lot of attention from developers over the weekend that described an alarming incident involving a company using Google's cloud.
- The post discussed an email received out of the blue from Google, warning that if the company didn't provide some specific forms, Google could decide to permanently shut down the service.
- The post has already been updated to say that Google has fixed the problem, and all appears to be well - but the post set off a conversation among developers over the quality of Google's customer service.
A post published to Medium on Thursday stirred a lot of debate about Google's ability - or, perhaps, inability - to provide adequate customer service to customers using its Google Cloud platform.
The situation has been resolved, and support people from Google's cloud have already responded, but not before some serious questions were raised about how Google Cloud does business.
The post was written by an anonymous administrator of a system that monitors "hundreds of wind turbines and scores of solar plants," it said. The admin wrote about receiving an email from Google on Thursday that said the company's website and other services were blocked because of "suspicious activity."
"All my systems have been turned off," wrote the anonymous administrator. "Everything is off. The machine has pulled the plug with no warning. The site is down, app engine, databases are unreachable, multiple Firebases say I've been downgraded and therefore exceeded limits."
Then came the real scare. The admin says they got a message saying that if the account owner didn't correct the violation by filling out the Account Verification Form and supply identification within three days, Google may decide to permanently close the account. That indicated to him and many of the people reading the Medium post that Google was willing to totally destroy an app critical to the business with only three days of notice.
"I hope (the Google Cloud Platform) team is listening and changes things for the better," the admin wrote. "Until then I'm never building any project on GCP."
Google did not respond to requests for comment and Business Insider could not verify the authenticity of the Medium post. Nonetheless, the problems described in the post are not unprecedented: In 2016, there was a similar case with a company named DocGraph.
Google needs more human service reps
In this case, everything appears to have worked out. The admin returned the filled-in form and the operation was back up in 20 minutes. A note was later added to the Medium post that said the Google Cloud support team had written and "assured us these incidents will not repeat."
But before the situation was resolved, the panicked admin had tried to reach Google customer service reps but couldn't raise anyone by phone or by chat.
And his post asked an important question: What would have happened if the person with the information Google required was away or otherwise out of touch for three days? Would Google have automatically followed through with its threat and shut the site down permanently?
The incident struck a chord with many developers, generating over 1,200 comments on message boards like Reddit and Hacker News about the quality of Google's cloud versus Amazon's. What many people seemed to agree on is that GCP needs to rely less on automated systems and provide more human support.
The lack of living, breathing customer-service help has long been a knock across many of Google's gadgets and services.
"This post is not about the quality of Google Cloud products," the admin wrote. "They are excellent, on par with AWS. This is about the no-warnings-given, abrupt way they pull the plug on your entire systems if they (or the machines) believe something is wrong. This is the second time this has happened to us."
In 2016, Fred Trotter, CEO of a healthcare startup called DocGraph, found himself in a similar situation. Trotter discovered that Google blocked access to the data he stored in the cloud after claiming one of his servers was involved in unspecified "intrusion attempts against a third-party."
Trotter tried to get information but again, his attempts to locate a human at Google Cloud were unsuccessful. Turns out the server in question had been hacked and was indeed used to launch attacks. And just like the recent case, Google then moved quickly to get Trotter's service working again.Steve Jennings/Getty
Google self sabotaging itself
Google is trying to take on Amazon in the burgeoning cloud computing market. If successful, this is a huge growth area for Google. In 2015, Google's Urs Hölzle even predicted that Google may one day make more revenue on cloud than on ads.
As it stands, Amazon's AWS is the clear leader, with Microsoft's Azure in second place, but Google has high hopes for the division. That's why the board hired Diane Greene to run its cloud business, buying out her still-in-stealth startup $380 million in 2016 to nab her and adding her to its board of directors.
Greene, who made her name as a cofounder of VMware, has a strong reputation in the enterprise IT world and has already done a lot to bring Google the kind of human support it needs to win corporate customers.
Often in the past, she has claimed that Google is cutting into Amazon's lead and possesses the better technology.
But technology aside, by not offering customer-service reps for emergencies for all levels of developers, or by using automated systems that fail to adequately address a problem, Google sabotages its efforts to overtake Amazon.
Interestingly enough, the author of the original Medium post writes that a Google Cloud engineer added in a private message that "there is a large group of folk (within Google Cloud Platform) interested in making things better, not just for you but for all GCP."
That would seem to confirm there's a problem.
On Twitter, Gartner analyst Lydia Leong put a fine point on the matter in response to this incident:
"GCP is for businesses. There is no excuse for this kind of thing. Not a 'cloud' problem. AWS and Azure don't do this," Leong wrote.