I'm not saying it's the case here, but sometimes the cloud providers themselves have inter-zone dependencies, or even inter-region dependencies that can cause outages in other zones and regions.
honestly more often than not the customer has a weird setup that doesn't handle it more than it being a GCP issue. I've seen some customers with awesomely redundant systems with their...cache all in one zone.
Also keep in mind GCP has a global network so it's possible some other regions are seeing disconnects/packet drops, though the source of the issue is contained to a single datacenter.
If you mean Google: Yes, extensive. One of the techniques they use is DiRT (Disaster Recovery Testing) which is a scheduled time-frame in which Google purposefully breaks systems that are exceeding their SLAs to verify that automated recovery executes as intended. In general, a DiRT test would catch something like "This service has a key dependency on us-east-1."
Now, in the past (also a Xoogler)... Some pieces of Cloud were moving at a cadence that they got away with a bit of a relaxed DiRT awhile back, so if someone tells me "I remember when a us-east-1 outage took down a key Cloud service" I would believe it. DiRT is also planned testing, it's not slapping a chaos monkey on the cluster fabric and crossing one's fingers, so it's entirely possible for key outage modalities to get missed (engineers are human and even though Google aims for a "no haunted closets" engineering practice, people develop a sixth sense of where the weak parts are of your infra and it takes courage to commit to pushing on those until they break).
People who don't work on cloud services are usually very sloppy with their wording which I think is leading to some disconnect here. A "zone" is not really the unit of availability for most things people consider the "cloud" for hyperscalers. A zone is like 1-3 DCs and you are very vulnerable to stuff like localized power events. For that reason most services are not offered at the zonal level and if they are you get big disclaimers saying so and lower 9s of SLA.
Most things are at a regional level which is multiple zones. Your AWS us-east-1 is one such region. A lot of datacenters spread across a decently large physical area have to have an issue for a whole region to start having failures.
For GCP it seems like one power event in a DC led to a thundering herd on the rest of the region which was the actual issue here. (which is a pretty significant problem and isn't supposed to happen)
You are generally right, except that you don't seem to know the history of major AWS outages. Failures in AWS in us-east-1, which is the first AWS cloud region, often cause cascading worldwide failures in the whole of AWS - especially failures related to DNS and to DynamoDB, if I remember correctly. That is, AWS workloada running in, say, eu-west-1 (and in fact in the whole of AWS) often start failing or at least having greatly degraded performance/availability because of a failure in AWS infra in their core region.
If AWS themselves can't handle cross region isolation and availability well, my expectations for most companies running on top of AWS are quite modest.
It's because AWS runs their control plane out of us-east-1, particularly lambda. There are also some very popular services that can only be run out of us-east-1. It's not really a technological constraint, it's a design decision.
AWS does not run “their control plane” out of us-east-1. Every single service has its own control plane independently hosted in its own region, with the exceptions of IAM, Route53, Cloudfront, and a small number of others like Global Accelerator.
IAM is usually the culprit when us-east-1 issues affect other regions, but Lambda runs independently in every region.
Explain why how why when aws lambda when out in us-east-1 a year or so ago every other region had intense issues they couldn’t quickly recover from? and their incident report explained this exact fact? where’s your info coming from? I follow this stuff very closely because my livelihood depends on it.
"While the event started at 11:48 PM PDT on October 19 and ended at 2:20 PM PDT on October 20, there were three distinct periods of impact to customer applications. First, between 11:48 PM on October 19 and 2:40 AM on October 20, Amazon DynamoDB experienced increased API error rates in the N. Virginia (us-east-1) Region. Second, between 5:30 AM and 2:09 PM on October 20, Network Load Balancer (NLB) experienced increased connection errors for some load balancers in the N. Virginia (us-east-1) Region. This was caused by health check failures in the NLB fleet, which resulted in increased connection errors on some NLBs. Third, between 2:25 AM and 10:36 AM on October 20, new EC2 instance launches failed and, while instance launches began to succeed from 10:37 AM, some newly launched instances experienced connectivity issues which were resolved by 1:50 PM."
At this point we’re playing semantic nittery. us-east-1 is their control plane. critical services that only run in us-east-1 were impacted badly. this has a cascading effect to services that run in other regions that rely on those services - even if those services themselves aren’t technically out, it’s a cascading effect. the whole fucking internet went down that day.
my original claim was that their control plane runs in us-east-1. this is a fact.
> my original claim was that their control plane runs in us-east-1. this is a fact.
The claim I'm testing is whether the impact spread to AWS services in other regions.
> At this point we’re playing semantic nittery.
It's not "semantic nittery." It's important as an engineer to be precise about the impact of incidents and how systems are architected, as opposed to making broadly sweeping (and false) claims that "[an entire] control plane runs in us-east-1" and "the whole fucking internet went down that day."
As someone who worked for AWS, I have insider knowledge that what you said is not true as a blanket statement. They generally take great care to isolate both control and data planes regionally, with rare exceptions where service data has to be globally consistent. Moreover, they have always recommended not to rely on control planes to respond to incidents: https://docs.aws.amazon.com/wellarchitected/latest/framework...
> For GCP it seems like one power event in a DC led to a thundering herd on the rest of the region which was the actual issue here. (which is a pretty significant problem and isn't supposed to happen)
Why do you say it's not supposed to happen? It's the predictable outcome of not having unlimited scale. The hyperscalers are big, but to be able to accommodate a tectonic shift in demand in a very compressed amount of time is very costly--costs that will be passed to their customers--if it can be done at all given current constraints. Today's hardware crunch, even for hyperscalers, is very real; customers are hoovering up spare capacity like never before.
As a former guidance counselor (specialist SA) for AWS, I regularly told customers that they shouldn't expect to be able to pack up and relocate into a new AZ once an incident commenced. They rarely listened, as it was often deemed too costly to maintain AZ redundancy (let alone regional redundancy) in advance of an incident.
> They rarely listened, as it was often deemed too costly to maintain AZ redundancy (let alone regional redundancy) in advance of an incident.
Do you blame them? You not only have to pay double for the infra but also be charged for inter-region data transfers. And, when the worst comes, you realize it was all for nothing as IAM doesn't work, for example.
> You not only have to pay double for the infra but also be charged for inter-region data transfers.
Well, yeah. Redundancy isn't free. I don't see how this is controversial.
> And, when the worst comes, you realize it was all for nothing as IAM doesn't work, for example.
What do you mean, "IAM doesn't work"? IAM has always worked, even during control plane impairments. You cannot depend on control plane operations to work during incidents; you should plan in advance for failure and have your orchestration already set up so that you don't have to touch a thing when you have to weigh away from an AZ or region.
This points to what I keep repeating about the control plane vs. the data plane. IAM faulted to a steady state. While you couldn’t make changes during the impairment, it didn’t cause downstream failures like blocking rule evaluations (or worse, failing open and allowing denied operations to be permitted). That’s the important part.
As the page you linked to stated:
> IAM (Identity and Access Management) configuration was unavailable, preventing updates to roles and policies, but cached credentials allowed most running workloads to continue.
Perhaps I should have been clearer about what I meant about IAM still working. That's on me.
This is why Microsoft calls it an Availability Zone, each zone is made up of one or more data centers equipped with entirely independent power, cooling, and networking infrastructure. One zone should not knock out other zones. Zones are also often nearby each other just enough that they have very efficient fiber optic connections between them.
That said, that's how Microsoft handles it, I'm not sure how GCP compares.
It is an interesting conundrum though, what if all your nearby availability zones come down at once due to power loss? I wonder which major cloud provider has the best setup for these sort of events.
GCP has some regions where all of the AZs are located in the same physical building. AFAIK they’ve stopped doing this in newer regions and have been migrating existing regions to have physically separate sites for each AZ.
AWS has always guaranteed that different AZs are physically separate sites, but that hasn’t always been the case with GCP.
>It is an interesting conundrum though, what if all your nearby availability zones come down at once due to power loss? I wonder which major cloud provider has the best setup for these sort of events
im only familiar with microsoft in this sense, but they have "geo-zone-redundant storage" which copies data across 3+ availability zones in the primary region, and also a secondary geographic region (i think at least 100 miles away if i remember, but maybe further). any incident would have to be massive to affect multiple geo-zones
They had a comparable problem in europe-west9 in 2023. Apparently they actually run into trouble if they lose a whole zone through a physical event. It's either thundering herd or simply insufficient capacity in the other two zones to take over all the load from the third zone.
1. AWS has no cross-regional or cross-zonal data plane[1] dependencies. Customers might couple them, in violation of best practices, but AWS itself does not. It's strictly forbidden.
2. That article you linked to does not support your claim. That is a report of a single-zone incident impacting us-east-1 only. Five other AZs in the same region were still fully operational.
[1] These are not the same as control plane dependencies: you might not be able to make certain mutations, but everything should still work in the steady state.
While what you say is generally true for 1, I wholly disagree that it is always customers that do the coupling.
One more than one occasion there are AWS URLs, including redirects inside the console, that include US East 1 in the URL that were only spotted during an outage. Manually changing it to an different region allowed us to get around it but to wholely blame customers when you don't know the context is bad form.
But there are also always exceptions. I've seen US East 1 outages cause autoscaling issues in other regions including failing to start new EC2s. A real problem if things have scaled down overnight.
> One more than one occasion there are AWS URLs, including redirects inside the console, that include US East 1 in the URL that were only spotted during an outage. Manually changing it to an different region allowed us to get around it
Yeah, that tracks. The Console has often been wonky like that.
> I've seen US East 1 outages cause autoscaling issues in other regions including failing to start new EC2s
Do you mean scaling failed due to insufficient capacity, or something else? ICE errors I get (thundering herd of attempted capacity shifting); but otherwise, that sounds implausible, and I respectfully believe you are mistaken about cause and effect. Happy to evaluate a causal chain of evidence that proves you correct, though.
My documentation for events from 2018 at this point are limited, and trying to go back through the lists of events demonstrates the limited reporting for some things deemed "non-critical" such as the early 2019 connectivity issues with EU West 2 when the London fibres were cut and services such as Workspaces experienced an outage.
I will do better at saving exact details for internet strangers in future.
2018 was almost a decade ago. Things have changed a lot since then. If you'd included that fact up front, I would have been more likely to believe you.
Actually in the beginning it tried to be a community but they forced it to be only an answer machine. They created and enforced a rule against anything outside of the strict Q&A format.
It's been years since, but I too had brain zaps coming off of it. It took a few months for those to go away.
I tried to describe it to my doctor and he seemed like he never heard of it before and sort of look at me like I was crazy, or at least that is what I sensed. But I ended up researching online and found it was a thing that happens.
Interesting is your doctor a general practitioner or a psychiatrist? A psychiatrist definitely knows about this stuff. GPs in my experience know almost nothing about anything beyond what seems like a script they operate from.
A GP nowadays is a triage person and nothing more. Sadly. My old school GP back in the day was a proper doctor. My current one is a really really nice guy who knows less about anything than I do. I literally tell him what to prescribe me and he doesn't know what it is.
Whenever I have some issue he never diagnoses it but sends me off to radiology because those doctors will come up with a diagnosis for him.
He then might treat if simple or more likely triage me to see a specialist.
I kind of find it cute to be able to be paid for this. I could do it easily and better but I would have to spend years studying medicine, which at 50 is not an option anymore
(this is just to my modern GP's job. I could definitely today NOT do my old proper full-on-doctor GP's job)
A psychiatrist definitely does not always know these things. Mine told me there should be no Citalopram withdrawals and that I should taper by cutting my dose in half over a 2 week period.
They were absolutely wrong as I have serious withdrawal reactions to Citalopram. It took me a month of micro-dosing down to safely get off a drug I was on for 15 yrs. And even then I was having constant brain & body zaps, that feel like short circuiting, for a full month after I dosed down to nothing. 6 mos later & I still can't sleep more than 6 hrs.
How did I know this was going to happen? Going cold turkey when my script ran out. I even told my psych about it numerous times but they just brushed it off.
Honestly, this is where chatbots and AI come into play and make an astonishing difference. I’m an AI realist/pessimist but its usefulness in general medicine is going to save millions of lives due to doctor ineptitude. They are after all just well read mechanics
This guy should get his thyroid checked.