Nothing can have an unlimited uptime. But even when AWS and other hyperscalers are having incidents, we've created an architecture that is resilient to those outages. Here's how we do it for Authress.
One of the most massive AWS incidents transpired on October 20th. The long story short is that the DNS for DynamoDB was impacted for us-east-1, which created a health event for the entire region. It's the worst incident we've seen in a decade. Disney+, Lyft, McDonald'ss, New York Times, Reddit, and the list goes on were lining up to claim their share too of the spotlight. And we've been watching because our product is part of our customers critical infrastructure. This one graph of the event says it all:
Murphy's Law
Everything fails all the time. There absolutely will be failures everywhere. Every line of code, every component you pull in, every library, there's guaranteed to be a problem in each and everyone of those. And you will for sure have to deal with it, at some point. So being prepared to handle that situation, is something you have to be thinking through in your design.
DNS
DNS, yeah, AWS will say it, everyone out there will say, and now we get to say it. The global DNS architecture is pretty good and reliable for a lot of scenarios, but I worry that it's still a single point of failure in a lot of ways.
Infrastructure as Code (IAC)
The last thing is infrastructure as code challenges. We deploy primary regions, but then there's also the backup regions, which are slightly different from the primary regions, and then there are edge compute, which are, again, even more slightly different. And then sometimes, we do this ridiculous thing, where we deploy infrastructure dedicated to one customers. And in doing so, we're running some sort of IaC to deploy those resources.
It is almost exactly the same architecture. Almost! Because it isn't exactly the same there are quite the opportunities for challenges to sneak it. That's problematic with even Open Tofu or CloudFormation, and often these tools make it more difficult, not less. And good luck to you, if you're still using some else that hasn't been modernized. With those, it's even easier to run into problems and not get it exactly correct.
The last thing I want to leave you with is, well, With all of these, is that actually sufficient to achieve five nines?
No. Our commitment is 5-nines, what we do is in defense of that, just because you do all these things doesn't automatically mean your promise of 5-nines in guaranteed. And you know what, you too can promise a 5-nines SLA without doing anything. You'll likely break your promise, but for us our promise is important, and so this is our defense.
Read the original article
Comments
I'm surprised the section about retries doesn't mention correlations. They say:
> P_{total}(Success) = 1 - P_{3rdParty}(Failure)^{RetryCount}
By treating P_{3rdParty}(Failure) as fixed, they're assuming a model in which each each try is completely independent: all the failures are due to background noise. But that's totally wrong, as shown by the existence of big outages like the one they're describing, and not consistent with the way they describe outages in terms of time they are down (rather than purely fraction of requests).
In reality, additional retries don't improve reliability as much as that formula says. Given that request 1 failed, request 2 (sent immediately afterward with the same body) probably will too. And there's another important effect: overload. During a major outage, retries often decrease reliability in aggregate—maybe retrying one request makes it more likely to go through, but retrying all the requests causes significant overload, often decreasing the total number of successes.
I think this correlation is a much bigger factor than "the reliability of that retry handler" that they go into instead. Not sure what they mean there anyway—if the retry handler is just a loop within the calling code, calling out its reliability separately from the rest of the calling code seems strange to me. Maybe they're talking about an external queue (SQS and the like) for deferred retries, but that brings in a whole different assumption that they're talking about something that can be processed asynchronously. I don't see that mentioned, and it seems inconsistent with the description of these requests as on the critical path for their customers. Or maybe they're talking about hitting a "circuit breaker" that prevents excessive retries—which is a good practice due to the correlation I mentioned above, but if so it seems strange to describe it so obliquely, and again strange to describe its reliability as an inherent/independent thing, rather than a property of the service being called.
Additionally, a big pet peeve of mine is talking about reliability without involving latency. In practice, there's only so long your client is willing to wait for the request to succeed. If say that's 1 second, and you're waiting 500 ms for an outbound request before timing out and retrying, you can't even quite make it to 2 full (sequential) tries. You can hedge (wait a bit then send a second request in parallel) for many types of requests, but that also worsens the math on overload and correlated failures.
The rest of the article might be much clearer, but I have a fever and didn't make it through.
> the section about retries doesn't mention correlations. [...] By treating P_{3rdParty}(Failure) as fixed, they're assuming a model in which each each try is completely independent: all the failures are due to background noise. But that's totally wrong, as shown by the existence of big outages like the one they're describing
Yes, that jumped out at me as well. A slightly more sophisticated model could be to assume there are two possible causes of a failed 3rd party call: (a) a transient issue - failure can be masked by retrying, and (b) a serious outage - where retrying is likely to find that the 3rd party dependency is still unavailable.
Our probabilistic model of this 3rd party dependency could then look something like
I.e. a failed call is 9x more likely to be caused by a transient issue than a serious outage. If the cause was a transient issue we assume independence between sequential attempts like in the article, but if the failure was caused by a serious outage there's only a 5% chance that each sequential retry attempt will succeed.P(first call failure) = 0.10 P(transient issue | first call failure) = 0.90 P(serious outage | first call failure) = 0.10 P(call failure | transient issue, prior call failure) = 0.10 P(call failure | serious outage, prior call failure) = 0.95In contrast with the math sketched in the article, where retrying a 3rd party call with a 10% failure rate 5 times could suffice for a 99.999% success rate, with the above model of failure modes including a serious outage failure mode producing a string of failures, we'd need to retry 135 times after a first failed call to achieve the same 99.999% success rate.
Your points about overall latency client is willing to wait & retries causing additional load are good, in many systems "135 retry attempts" is impractical and would mean "our overall system has failed and is unavailable".
Anyhow, it's still an interesting article. The meat of the argument and logic about 3rd party deps needing to meet some minimum bar of availability to be included still makes sense, but if our failure model considers failure modes like lengthy outages that can cause correlated failure patterns, that raises the bar for how reliable any given 3rd party dep needs to be even further.
This is absolutely true, but the end result is the same. The assumption is "We can fix a third party component behaving temporarily incorrectly, and therefore we can do something about it". If the third party component never behaves correctly, then nothing we can do to fix it.
Correlations don't have to be talked about, because they don't increase the likelihood for success, but rather the likihood of failure, meaning that we would need orders of magnitude more reliable technology to solve that problem.
In reality, those sorts of failures aren't usually temporary, but rather systemic, such as "we've made an incorrect assumption about how that technology works" - feature not a bug.
In that case, it doesn't really fit into this model. There are certainly things that would better indicate to us that we could use or are not allowed to use a component, but for the sake of the article, I think that was probably going much to far.
TL;DR Yes for sure, individual attempts are correlated, but in most cases, it doesn't make sense to track that because those situations end up in other buckets of "always down = unreliable" or "actually up - more complex story which may not need to be modelled".
I think the reasoning matters as much as the answer, and you had to make at least a couple strange turns to get the "right answer" that retries don't solve the problem:
* the 3rd-party component offering only 90% success—I've never actually seen a system that bad. 99.9% success SLA is kind of the minimum, and in practice any system that has acceptable mean and/or 99%/99.9% latency for a critical auth path also has >=99.99% success in good conditions (even if they don't promise refunds based on that).
* the whole "really reliable retry handler" thing—as mentioned in my first comment, I don't understand what you were getting at here.
I would go a whole other way with this section—more realistic, much shorter. Let's say you want to offer 99.999% success within 1 second, and the third-party component offers 99.9% success per try. Then two tries gives you 99.9999% success if the failures are all uncorrelated but retries do not help at all when the third-party system is down for minutes or hours at a time. [1] Thus, you need to involve an alternative that is believed to be independent of the faulty system—and the primary tool AWS gives you for that is regional independence. This sets up the talk about regional failover much more quickly and with less head-scratching. I probably would have made it through the whole article yesterday even in my feverish state.
[1] unless this request can be done asynchronously, arbitrarily later, in which case the whole chain of thought afterward goes a different way.
By wparad 2025-11-1822:32 Hmm, I never considered potentially using an SLA on latency as a potential way to justify the argument. If I pull this content into a future article or talk, I will definitely consider reframing it for easier understanding.
Agreed, I think the introduction is wrong and detracts from the rest of the article.
By wparad 2025-11-188:59 Hmmm, which part of the intro did you find an issue with? I want to see if I can fix it.
This is probably one of the best summarizations of the past 10 years of my career in SRE. Once your systems get complex enough, something is always broken and you have to prepare for that. Detection & response become just as critical as pre-deploy testing.
I do worry about all the automation being another failure point, along with the IaC stuff. That is all software too! How do you update that safely? It's turtles all the way down!
By wparad 2025-11-1721:52 Thank you!
One of the question I frequently get is "do you automatically rollback". And I have hide in the corner and say "not really". Often, if you knew a rollback would work, you probably could also have known to not roll out in the first place. I've seen a lot of failures that only got worse when automation attempted to turn the thing on and off again.
Luckily from an automation roll-out standpoint, it's not that much harder to test in isolation. The harder parts to validate are things like "Does a Route 53 Failover Record really work in practice at the moment we actually need it to work?"
Usually the answer is yes, but then there's always the "but it too could be broken", and as you said, it's turtles all the way down.
The nice part is realistically, the automation for dealing with rollout and IaC is small and simple. We've split up our infrastructure to go with individual services, so each piece of infra is also straight forward.
In practice, our infra is less DRY and more repeated, which has the benefit of avoiding complexity that often comes from attempting to reduce code duplication. The ancillary benefit is that, simple stuff changes less frequently. Less frequent changes because less opportunity for issues.
Not-surprisingly, most incidents comes from changes humans make. Where the second most amount of incidents come from assumptions humans make about how a system operates in edge conditions. If you know these two things to be 100% true, you spend more time designing simple systems and attempting to avoid making changes as much as possible, unless it is absolutely required.
Iac is definitely a failure point, but the manual alternative is much worse! I’ve had a lot of benefit from using pulumi, simply because the code can be more compact than the terraform hcl was.
For example, for the fall over regions (from the article) you could make a pulumi function that parameterizes only the n things that are different per fall over env and guarantee / verify the scripts are nearly identical. Of course, many people use modules / terragrunt for similar reasons, but it ends up being quite powerful.
By wparad 2025-11-1721:57 I think some people are going to scream when I say this, but we're using mostly CloudFormation templates.
We don't use the CDK because it introduces complexity into the system.
However to make CloudFormation usable, it is written in typescript, and generates the templates on the fly. I know that sounds like the CDK, but given the size of our stacks, adding an additional technology in, doesn't make things simpler, and there is a lot of waste that can be removed, by using a software language rather than using json/yaml.
There are cases we have some OpenTofu, but for infrastructure resources that customer specific, we have deployments that are run in typescript using the AWS SDK for javascript.
It would be nice if we could make a single change and have it roll-out everywhere. But the reality is that there are many more states in play then what is represented by a single state file. Especially when it comes to interactions between—our infra, our customer's configuration, and the history of requests to change the configuration, as well as resources with mutable states.
One example of that is AWS certificates. They expire. We need them expiring. But expiring certs don't magically update state files or stacks. It's really bad to make assumptions about a customer's environment based on what we thought we knew the last time a change was rolled out.
I actually like terraform for its LACK of power (tho yeah these days when I have a choice I use a lot of small states and orchestrate with tg).
Pulumi or CDK are for sure more powerful (and great tools) but when I need to reach for them I also worry that the infra might be getting too complex.
By yearolinuxdsktp 2025-11-1722:143 reply IMO Pulumi and CDK are an opportunity to simplify your infra by capturing what you’re working with using higher-level abstractions and by allowing you to refactor and extract reusable pieces at any level. You can drive infra definitions easily from typed data structures, you can add conditionals using natural language syntax, and stop trying to program in a configuration language (Terraform HCL with surprises like non-short-circuited AND evaluation).
You still end up having IaaC. You can still have a declarative infrastructure.
By andrewaylett 2025-11-1722:52 That's how we use CDK. Our CDK (in general) creates CloudFormation which we then deploy. As far as the tooling which we have for IaC is concerned, it's indistinguishable from hand-written CloudFormation — but we're able to declare our intent at a higher level of abstraction.
By darkwater 2025-11-1811:17 > and stop trying to program in a configuration language
Many people don't program with a configuration language like HCL. We use it as what it is - a DSL - that covers its main use case in an elegant manner. Maybe I never touched complex enough infra that twists a DSL into a general-use language, but in my experience there are simply no real benefits when using something like CDK (I never tried Pulumi to be fair).
By xyzzy123 2025-11-1722:44 Absolutely, the best case is it's much better, safer, readable etc. However, the worst case is also worse. From the perspective of someone who provides devops support to multiple teams, terraform is more "predictable".
By wparad 2025-11-1721:58 Agreed, it is much too easy to fall into bad habits. The whole goal of OpenTofu is declarative infrastructure. With CDK and pulumi, it's very easy to end up in a place where you lose that.
But if you need to do something in a particular way, the tools should never be an obstacle.
By spyspy 2025-11-1721:40 If you do use terraform, for the love of god do NOT use Terraform Cloud. Up there with Github in the list of least reliable cloud vendors. I always have a "break glass" method of deploying from my work machine for that very reason.
By sharklasers123 2025-11-1719:093 reply Is there not an inherent risk using an AWS service (Route 53) to do the health check? Wouldn’t it make more sense to use a different cloud provider for redundancy?
By kondro 2025-11-180:46 While there appears to be some us-east-1 SPoF for Route 53 updates (as shown recently), the actual health checks themselves occur in up to 8 different regions [1] with an 18%[2] agreement of failure required to initiate a failover.
AWS has very good isolation between regions and, while it relies on us-east-1 for control plane updates to Route 53, health checks and failovers are data plane operations[3] and aren't affected by a us-east-1 outage.
Relying on a single provider always seems like a risk, but the increased complexity of designing systems for multi-cloud will usually result in an increased risk of failure, not a decrease.
1. us-east-1, us-west-1, us-west-2, eu-west-1, ap-southeast-1, ap-southeast-2, ap-northeast-1 and sa-east-1 which defaults to all of them.
2. https://docs.aws.amazon.com/Route53/latest/DeveloperGuide/dn...
3. https://aws.amazon.com/blogs/networking-and-content-delivery...
If the check can't be done, then everything stays stable, so I'm guessing the question is, "What happens if Route 53 does the check and incorrectly reports the result?"
In that case, no matter what we are using there is going to be a critical issue. I think the best I could suggest at that point would be to have records in your zone that round robin different cloud providers, but that comes with its own challenges.
I believe there are some articles sitting around regarding how AWS plans for failure and the fallback mechanism actually reduces load on the system rather than makes it worse. I think it would require in-depth investigation on the expected failover mode to have a good answer there.
For instance, just to make it more concrete, what sort of failure mode are you expecting to happen with the Route 53 health check? Depending on that there could be different recommendations.
By indigodaddy 2025-11-1721:161 reply Have you considered the scenario of "everything is so dead in aws", that the check doesn't happen, plus the backends are dead too (this is assuming the backend services live in aws as well) ? But I'd guess in that case you'd know quickly enough from supplementary alerting (you guys don't seem the type to not have some sort of awesome monitoring in place) and you have a different/worse DR problem on your hands.
As far as the OP's point though, I'm going to probably assume that the health checks need to stay within/from AWS because 3rd party health checks could taint/dilute the point of the in-house AWS HC service to begin with.
I think there are two worlds of thought to the "AWS is totally dead everywhere". And that's: * It is never going to happen due to the way AWS is designed (or at least told to us, which explains why it is so hard to execute actions across regions.) * It will happen but then everything else is going to be dead, so what's the point?
One problem we've run into, which is the "DNS is single point of failure" is that there isn't a clear best strategy to deal with "failover to a different cloud at the DNS routing level."
I'm not the foremost expert when it comes to ASNs and BGPs, but from my understanding that would require some multi-cloud collaboration to get multiple CDNs to still resolve, something that feels like it would require both multiple levels of physical infrastructure as well as significant cost to actually implement correctly compared to the ROI for our customers.
There's a corollary here for me, which is, still as simple as possible to achieve the result. Maybe there is a multi-cloud strategy, but the strategies I've seen still rely on having the DNS zone in one provider that fail-overs or round-robins specific infra in specific locations.
Third party health checks have less of a problem of "tainting" and more just cause further complications, as you add in complexity to resolving your real state, the harder it is to get it right.
For instance, one thing we keep going back and forth on is "After the incident is over, is there a way for us to stay failed-over and not automatically fail back".
And the answer for us so far is "not really". There are a lot of bad options, which all could have catastrophic impacts if we don't get it exactly correct, and haven't come with significant benefits, yet. But I like to think I have an open mind here.
By parliament32 2025-11-180:39 There is good options if you're willing to pay for them, but they have nothing to do with DNS. You will never get DNS TTLs low enough (and respected) to prevent a multi-minute service interruption in cases like these.
Proper HA is owning your own IP space and anycast advertising it from multiple IXes/colos/clouds to multiple upstreams / backbone networks. BGP hold times are like a dead-mans-switch and will ensure traffic stops being routed in that direction within a few seconds in case of a total outage, plus your own health-automation should disable those advertisements when certain things happen. Of course, you need to deal with the engineering complexity of your traffic coming in to multiple POPs at once, and it won't be cheap at all (to start, you're looking at ~10kUSD capex for a /24 of IP space, plus whatever the upstreams charge you monthly), but it will be very resilient to pretty much any single point of failure, including AWS disappearing entirely.
By toast0 2025-11-1722:43 It's painful, but you can split your DNS across multiple providers. It's not usually done other than during migrations, but if you put two NS names from providerA and two from providerB, you'll get a mix of resolution (most high profile domains have 4 NS names; sometimes based on research/testing, sometimes based on cargo culting; I assume you want to fit in... but amazon.com has 8, and the DNS root and some high profile tlds have 13, so you do you :)). If either provider fails and stops responding, most resolvers will use the other provider. If one provider fails and returns bad data (including errors) or even can no longer be updated [1], the redundancy doesn't really help --- you probably went from a full outage that's easy to diagnose to a partial outage that's much harder to diagnose; and if both providers are equally reliable, you increased your chances of having an outage.
[1] But, it's DNS; the expectation is that some resolvers, hopefully very few of them, will cache data as if your TTL value was measured in days. IMHO, If you want to move all your traffic in a defined timeframe, DNS is not sufficient.
By indigodaddy 2025-11-1719:351 reply Had the same thought, eg if things are really down can it even do the check etc
By hnuser123456 2025-11-1721:03 Ask some friends and family if you can install an RPi on their home network that monitors your service.
