On November 15, 2026, AWS reported connectivity issues in us-east-1 from 3:05 AM to 6:21 AM PST that affected 11 services, including IAM Identity Center sign-in and Elastic Load Balancing, where a subset of Network Load Balancers in a single cell of use1-az1 was affected. AWS has not published a root cause. The design lessons are to keep a break-glass sign-in path that does not depend on single sign-on, to spread load balancer targets across multiple Availability Zones, and to keep workloads already running in a second region.

Disclosure: Control Plane publishes this guide.

TL;DR

On Sunday, November 15, 2026, AWS us-east-1 had increased connectivity issues for 3 hours and 16 minutes, from the start time AWS gave (3:05 AM PST) to the last recovery (6:21 AM PST), according to the AWS Health Dashboard.

  • IAM Identity Center and AWS Sign Up were among 11 affected services, which also included Elastic Load Balancing, where a subset of Network Load Balancers (NLBs) in a single cell of use1-az1 was affected.
  • AWS fully mitigated the authentication issue at 5:20 AM PST. The NLB subset recovered last, at 6:21 AM PST.
  • AWS’s updates do not mention a cyberattack.
  • When IAM Identity Center sign-in fails, responders lose SSO access to the console and CLI, which slows incident response.
  • NLB nodes route to targets in their own Availability Zone by default, so targets in a single zone fail with that zone.
  • A second region that already serves traffic keeps users served when one region is impaired.
  • On Control Plane, a workload runs in every location of its Global Virtual Cloud (GVC), Control Plane’s set of locations across clouds, so when it runs in two or more locations and one fails, users are routed to the healthy ones.

What happened in AWS us-east-1 on November 15, 2026?

AWS gave these times in its resolved update on the AWS Health Dashboard, posted at 6:53 AM PST.

Time (PST) What AWS reported
3:05 AM Start of signup errors, authentication issues, and connectivity issues in us-east-1, per AWS’s resolved update. AWS lists 11 affected services, including IAM Identity Center and AWS Sign Up.
4:38 AM AWS posts its first update.
5:20 AM AWS fully mitigated the issue affecting authentication, and most AWS services observed full recovery.
6:10 AM End of the event window given in AWS’s resolved update.
6:21 AM AWS fully mitigated the underlying issue for a subset of customers using NLB in a single cell in use1-az1.
6:53 AM AWS posts its resolved update.

AWS’s resolved update gives 3:05 AM to 6:10 AM as the event window, and the same update says a subset of customers using NLBs in a single cell of use1-az1 had connectivity issues until 6:21 AM, 11 minutes later.

The 11 services AWS listed are IoT Device Management, CloudWatch, Connect, OpenSearch, Elastic Load Balancing, IAM Identity Center, Kiro, NAT Gateway, Sign Up, Step Functions, and VPC Lattice.

Why did a sign-in outage matter so much?

IAM Identity Center is the sign-in path for many organizations that use AWS single sign-on. When it is unavailable, engineers cannot use SSO to open the AWS Management Console. CLI sessions that depend on SSO-issued credentials cannot be refreshed. Credentials already issued keep working until they expire. The people who need to respond to an incident lose their usual way in.

An organization instance of IAM Identity Center is enabled in one primary Region. If that Region has a disruption, SSO access to accounts can be affected even when the workloads in those accounts run elsewhere. AWS supports replicating an organization instance to additional Regions so workforce users can keep reaching their AWS accounts, with permissions provisioned before the disruption, if the primary Region is disrupted. Replication needs a multi-Region customer managed KMS key and is not available with an Active Directory identity source. Management access and workload health are separate concerns, and each needs its own fallback.

AWS documentation recommends setting up emergency access to the console before a disruption occurs. One option is direct federation from your external identity provider to individual AWS accounts. Another is break-glass access that does not depend on that identity provider. AWS also recommends testing the configuration periodically, because you cannot create the required IAM roles if your access to create them is also disrupted.

On-call engineers often learn about an incident from alerts and then need to sign in to investigate. If sign-in fails, escalation steps that assume console access stall. Runbooks that fetch SSO tokens stall too.

This table maps each failure type to what it blocks and a workaround that does not depend on the failing component.

Failure What it blocks Workaround that does not depend on it
IAM Identity Center sign-in down New console and CLI sign-ins through SSO, so engineers cannot sign in to respond Emergency access through direct IAM federation, break-glass IAM users or roles with strong MFA, or a replicated IAM Identity Center instance in a second Region, set up and tested in advance
NLB cell in one AZ impaired Traffic to targets behind that NLB in that zone Targets registered in multiple AZs, or a new NLB with targets in multiple AZs as AWS advised
A regional service in us-east-1 impaired Anything homed only in that region Workloads already running in a second region, with DNS failover configured
Status and alerting hosted in the same region You cannot see or announce the incident Status page and alerting hosted outside the failing region

Authentication, load balancing, regional placement, and communication are four failure domains, and each needs a fallback that does not share the others’ dependencies.

What does a single-cell Network Load Balancer failure mean for your architecture?

AWS reported that a subset of customers using NLB in a single cell in use1-az1 had connectivity issues until 6:21 AM PST. Two AWS behaviors explain why that matters.

The first is how Availability Zones are named. An AZ ID such as use1-az1 identifies the same physical zone in every AWS account. Zone names such as us-east-1a are mapped randomly per account, so us-east-1a in one account may be a different physical zone from us-east-1a in another. To check whether your resources sat in the affected zone, compare AZ IDs. The VPC console shows the AZ ID for each subnet, and the AWS CLI describe-availability-zones command returns the mapping for your account.

The second is how Network Load Balancers route traffic. By default, each NLB node distributes traffic only across the registered targets in its own Availability Zone. Cross-zone load balancing is off by default for NLBs. You can turn it on at the load balancer level or per target group. AWS documentation states that an NLB is most effective when each enabled Availability Zone has at least one registered target.

The NLB has one IP address per enabled Availability Zone, and its DNS name resolves to those addresses. If the node in one zone is impaired and that zone holds your only targets, traffic to those targets stops. Healthy instances in other zones do not receive it unless they are registered and reachable through a working node.

AWS recommended that customers who needed immediate resolution create a new NLB with targets in multiple Availability Zones, and later recommended retrying any failed requests. Acting on that advice during an incident requires three things. You already have capacity running in other zones. You can update DNS records or upstream references quickly. You can sign in to make the change. The sign-in issue earlier in the same morning shows why the third point needs its own plan.

Register NLB targets in at least two Availability Zones. Decide per workload whether to turn on cross-zone load balancing. Watch target health and zonal health per zone so a zonal problem shows up before users report it. For an NLB with targets in more than one zone, you can turn on zonal shift and use it to move traffic away from one zone. Zonal shift is off by default.

How do you prepare for the next us-east-1 incident?

This checklist covers the six areas this incident touched, with a way to test each one.

Area What to have in place How to test it
Break-glass access Emergency access configured before any disruption, through direct IAM federation, break-glass IAM users or roles with strong MFA that do not depend on IAM Identity Center, or a replicated IAM Identity Center instance in a second Region Run a periodic sign-in drill using only the emergency path
Multi-AZ NLB targets and cross-zone settings Targets registered in two or more AZs for every NLB, with cross-zone load balancing set per workload Deregister targets in one AZ in a test environment and confirm traffic still reaches the others
A second region already serving traffic At least one more region running the workload and receiving production traffic Route part of real traffic to the second region on a schedule and confirm end-to-end behavior
DNS failover with health checks Route 53 health checks on endpoints in each region, with failover records and TTLs short enough for your recovery target Fail a health check on purpose and confirm DNS resolves to the healthy region
Status page and alerting outside the failing region Incident communication and alert pipelines hosted in another region or with a provider that does not depend on us-east-1 Send a test alert while simulating loss of us-east-1 and confirm it reaches on-call
Runbooks that do not need SSO Recovery steps that use the emergency access path and pre-provisioned resources Walk each runbook in a tabletop exercise without SSO

Route 53 health checks can monitor endpoints, and DNS failover records route traffic away from an endpoint that fails its check. Configure health checks and failover records in advance, because editing Route 53 records depends on Route 53’s management API in us-east-1. Failover only helps if the second region already has capacity and data it can serve. Schedule each drill in the checklist so every runbook has run at least once before an incident.

How does this compare with other cloud outages?

The November 15, 2026 incident lasted 3 hours and 16 minutes.

The cloud outage tracker records 63 major AWS, Azure, Google Cloud, and Cloudflare incidents since January 2020, with a median duration of 5 hours and 57 minutes. By that measure, this incident was shorter than the median major incident across those providers.

Regional incidents happen on every large cloud. The useful response is design work: separate the sign-in path from the workload path, spread capacity across zones, and run in more than one region before you need to. Designing systems that survive the inevitable covers those patterns in more depth.

How Control Plane fits

Control Plane handles Day 2 operations for you, including autoscaling, built-in observability, and security. A Global Virtual Cloud (GVC) is Control Plane’s named set of locations across AWS, GCP, Azure, and Oracle Cloud Infrastructure regions, plus Kubernetes clusters you already run and your own servers. A workload runs in every location of its GVC, and by default every location serves traffic. When a workload runs in two or more locations and one fails, DNS routes users to the nearest location where the workload passes its readiness probes.

During the October 19 to 20, 2025 AWS us-east-1 outage, which lasted about 14 and a half hours, 20% of Control Plane customers had AWS us-east-1 in their GVC. Their workloads were active and running in healthy locations within 10 seconds, with no customer downtime, according to Control Plane’s internal incident telemetry. Control Plane is PCI DSS Level 1, SOC 2 Type II, HIPAA, and GDPR compliant.

Control Plane runs the workloads you deploy on it. It does not change how third-party SaaS, identity providers, or other organizations’ networks behave, so the break-glass and sign-in steps for AWS access still apply. For more on regional failures, read how companies stay online during AWS us-east-1 outages and how to keep your app online during an AWS region outage.

FAQ

These answers cover sign-in, load balancing, and regional design after the November 15, 2026 us-east-1 incident.

What happened to AWS us-east-1 on November 15, 2026?

AWS reported increased connectivity issues in us-east-1 from 3:05 AM to 6:21 AM PST, according to the AWS Health Dashboard. The incident affected 11 services, including IAM Identity Center sign-in and Elastic Load Balancing, where a subset of Network Load Balancers in a single cell of use1-az1 was affected. That NLB subset recovered last.

Was the November 15 AWS outage a cyberattack?

AWS’s updates do not mention a cyberattack. AWS described the event as increased connectivity issues.

How do I sign in to AWS when IAM Identity Center is down?

Use an access path you set up before the disruption. AWS documents replicating IAM Identity Center to an additional Region, direct IAM federation from your external identity provider to individual accounts, and break-glass access that does not depend on that identity provider. AWS recommends testing the setup periodically.

What is a break-glass account in AWS?

A break-glass account is an emergency IAM user or role, protected with strong MFA, that you use when normal sign-in paths such as IAM Identity Center are unavailable. Keep its credentials stored securely, limit who can use it, and test it on a schedule.

Are Network Load Balancers multi-AZ?

An NLB can span multiple Availability Zones, but by default each NLB node routes only to targets in its own zone. Cross-zone load balancing is off by default. Register targets in every enabled zone so the loss of one zone does not stop traffic.

What does use1-az1 mean?

use1-az1 is an Availability Zone ID in us-east-1 that refers to the same physical zone in every AWS account. Zone names such as us-east-1a are mapped randomly per account. Compare AZ IDs to check whether your resources were in the affected zone.

How do companies stay online during AWS us-east-1 outages?

They run workloads in at least two regions and use DNS failover with health checks to move traffic to the healthy region. They also keep an emergency sign-in path that does not depend on a single identity service. Status pages and alerting run outside the region they monitor.

Did AWS publish a root cause for the November 15 outage?

No. As of November 18, 2026, AWS has not published a root cause for the November 15 us-east-1 incident.

How long did the November 15, 2026 AWS outage last?

It lasted 3 hours and 16 minutes, from the start time AWS gave (3:05 AM PST) to the last recovery (6:21 AM PST). AWS’s resolved update lists 3:05 AM to 6:10 AM as the event window and reports that the NLB subset in use1-az1 had connectivity issues until 6:21 AM.

Should I move all my workloads out of us-east-1?

No. Run critical workloads in us-east-1 and at least one other region at the same time, with DNS failover and an independent sign-in path. A disruption in one region then affects a share of capacity, and the rest keeps serving.

Talk to the Control Plane team about keeping workloads serving from more than one region, or sign up free to start.

Originally published on the Control Plane blog.