Digital Ocean won't let new customers create resources in four DCs, won't say why

Digital Ocean has cloud capacity issues and won't say what the problem is. The junior cloud company's availability matrix, which lists resources customers can deploy, was last updated on July 18 and offers the following warning about four datacenters: Due to limited capacity in AMS2, NYC2, SFO1, and SFO2, resource creation …

[Anonymous Coward]

Slow news day?

I remember those DigitalOcean data centers not being available about 5 years ago. I don't know why, but I suppose an expansion of those DCs would be too expensive or difficult due to architectural reasons. But does it really matter? Fof all of the affected DCs, there is at least one other available data center in the same city.

[ecofeco]

Re: Slow news day?

Sir, this is Arbey's.

[imanidiot]

Energy prices?

Seems to me like they might be trying to keep energy costs down at the most expensive locations.

[doublelayer]

Re: Energy prices?

In all cases, they have another datacenter in the affected city where you can start up new things. It's possible that, for example, Amsterdam DC number 1 is efficient and number 2 isn't, but it's not about larger geographic locations. In addition, they've only frozen one European DC and three in North America, but most energy prices have increased more in Europe.

I'm not really sure why one chooses AMS1 or AMS2, so maybe they had too many people selecting one and need to get customers to select the other one. Or they just don't want to take the risk of scaling because they're cautious about the future demand but they're already close to limits.

[Camilla Smythe]

Probably overloaded with port scanners.

Either that or all of their IP space has been blocked by the rest of the internet having been associated with port scanners.

[Anonymous Coward]

Re: Probably overloaded with port scanners.

from what i've seen, they're moving to AWS.

[VoiceOfTruth]

Re: Probably overloaded with port scanners.

Ha ha! I have to agree 100%. The only web traffic we ever saw from Digital Ocean was script kiddies looking for a generic list of vulnerabilities. Eventually we just blocked them at the IP level. It's not worth even proving a http response. Sad but true. Digitial Ocean should do more to *police* its customers.

[Anonymous Coward]

Re: Probably overloaded with port scanners.

Yup, their range is in our blacklist too.

[Robert Grant]

> To us, it sounds as though Digital Ocean is particularly sensitive on cost – trying to wring the most out of its dollars, more so than you'd expect from a cloud player.

You may be used to cloud players who got scale through their existing cash cow(s).

[Korev]

With AWS it's the other way around :)

[Robert Grant]

True! Now.

[RJenson]

This is normal. Each region, which is a physical location like NYC (New York City), has a number of data centers. For example New York has NYC1, NYC2 and NYC3. A data center is a physical space with racks of servers.

When a data center like NYC1 is "filled up" a NYC2 is created in the same region (New York City). When that is filled up, a new datacenter NYC3 is created, etc. Additional capacity is left in NYC1 and NYC2 for existing customer growth. If customers remove resources from those data centers, more resources in them can become available for new customers. But if existing customers don't stop using resources, those "filled" data centers become unavailable for new customers as new customers should be creating resources in the new data centers, instead.

The limited capacity in AMS2, NYC2, SFO1, and SFO2 is because those data centers are filled with existing customers, along with additional resources reserved for those customers.

However, in each region, there are data centers with additional resources for new customers. In Amsterdam, there is AMS3. In New York, there is NYC1 and NYC3. In San Francisco, there is SFO3.

A data center has racks of machines and has a physical footprint. So, an individual data center can't grow forever because it is limited by physical space. Additional capacity is reserved in those locations for existing customers. However new customers that want to create resources in Amsterdam should use AMS3, in New York should use NYC1 or NYC3 or SFO3 in San Francisco.

Existing data centers can also become available for new customers when new hardware is deployed that enables additional resources in the same physical footprint. Server upgrades can allow 10X or more customers in the same physical location if the servers replaced are older as newer ones are much more powerful. So, data centers that are "filled" right now can also open up for new users if/when that happens.

[runt row raggy]

i think the surprise that El reg is expressing is that digital ocean bills itself as cloud but the concerns you mention are physical data center concerns, with no abstraction.

[MachDiamond]

Why I don't like the Cloud, let me count the ways.

Yes, I can see instances where contracting for Cloud services is beneficial, but it can also be a way to lose a whole bunch of customers very quickly.

All of Digital Ocean's clients that are affected are in a position of having put their reliance on having those services. Perhaps they've started fresh with the set up or maybe shifted from their own iron to them and then deleted their own hardware. Now they are in a position where they can't even throw money at the problem. Until DO provides a solution, they're dead in the water. If they were running their own system, even as a fall-back, they could buy hardware and pay massive overtime and bring in outside help right away to be back up and running as fast as possible. What ever can go wrong and all that.

There needs to be a devil's advocate that's listened to in any company that is looking to outsource something or uses outside services on a regular basis. I did. There were machining companies I could send work to if my main vendor was backlogged. I had a list of metal supply companies. The list goes on and on. Every component and service had an alternate and some thought was put into what-if scenarios. You make a list of things that can go wrong, plan what you will do in those situations and put all of it in 3-ring binders (really, physical binders since some of the problems might prevent accessing anything online). Many companies have stock, fill-in-the-blank PR statements for a variety of circumstances. Instead of an off-the-cuff statement after an accident, there will be a fully considered disclosure that can be sent out that says and admits to pretty much nothing, but in very crafted prose.