Amazon packages pile up after AWS outage spawns delivery havoc (detroitnews.com)

Article summary

An Amazon Web Services (AWS) outage caused disruptions to Amazon's delivery operation, preventing drivers from getting routes or packages and shutting down communication between Amazon and its drivers. The outage, which began at 10 a.m. eastern time, affected multiple AWS services, including Amazon DynamoDB and Amazon Elastic Compute, and also impacted other companies that rely on AWS, such as Netflix and Disney+. The outage was reported to have declined by about half for AWS and two-thirds for Amazon by 1:45 p.m. AWS identified the cause of the issue and was working to fix it, directing customers to alternative servers in its western region.

Main themes

• AWS outage
• cloud infrastructure
• redundancy and failover
• disaster recovery
• complexity and operational confidence
• distributed systems

What commenters say

• Some commenters believe that hosting core infrastructure in us-east-1 is a bad idea due to its history of outages and technical issues.
• Others argue that AWS is a distributed system and should not have a single point of failure, but acknowledge that complexity can introduce unknown errors.
• There is a debate about the importance of multi-region redundancy, with some arguing it is necessary for high availability and others claiming it is not always feasible or cost-effective.
• Some commenters think that Amazon should take responsibility for the outage and provide more robust infrastructure, while others believe that customers should build their own redundancy and failover systems.
• The idea that AWS is a magical solution that can prevent all outages is disputed, with some arguing that failure domains and operational confidence are still necessary.
• Some commenters note that even Amazon does not always follow its own advice on building multi-region systems, and that region outages can still happen.
• The importance of planning for disaster recovery and having a business continuity plan is emphasized by some commenters, who note that even a multi-AZ deployment may not be enough to prevent outages.
• There is a discussion about the trade-offs between reliability, cost, and complexity in building cloud-based systems, with some arguing that it is not always necessary to build highly available systems.