{"resourceId":"modelbank-bedrock-us-west-2-degradation-20261106","versions":[{"version":"external-dde4b26de0b38b4b7856127040bdc8b38b069a9a3be83e495bb299ee7f755e8b","resource":{"id":"modelbank-bedrock-us-west-2-degradation-20261106","title":"ModelBank: Amazon Bedrock degraded about 2.4 hours on Nov 6 as US-WEST-2 EC2/EBS launch failures hit dependent services","organization":"ModelBank.ai","sector":"Independent AI service availability monitoring","geography":"AWS US-West-2 (Oregon) commercial Region; GovCloud not involved per source","publishedAt":"2026-11-06","publicationDate":"2026-11-06","eventDate":"2026-11-06","sourceName":"ModelBank outage archive","sourceLabel":"AWS Bedrock outage on November 6, 2026: what happened and how long it lasted","sourceUrl":"https://modelbank.ai/outages/bedrock-2026-11-06","evidenceClass":"independent-reporting","outcomeClass":"cautionary","topics":["infrastructure","operating-model"],"finding":"ModelBank reports Amazon Bedrock had degraded performance on November 6, 2026 between 22:51 and 01:12 UTC (about 2.4 hours; 4:51–7:12 PM CT), mapped to one official AWS incident, 'Elevated Launch Failures' in US-WEST-2. Quoted AWS updates attribute the issue to a subsystem supporting EBS volumes needed during EC2 instance launch, affecting all four Oregon AZs and services that instantiate new instances (for example DataSync and AWS Backup); AWS advised customers not to mutate running instances. The resolved notice gives 2:35–4:21 PM PST. ModelBank's own probes detected degradation for about one hour. It reports 100.00% measured availability over 30 days and seven days with an outage or degradation in the last 90 days.","sledRelevance":"SLED Bedrock workloads hosted in US-West-2 can be affected by underlying EC2/EBS events; reinforces cross-Region or GovCloud cross-Region inference and resilience planning for constituent-facing AI.","evidence":"Third-party monitoring page combining its probes with quoted AWS status text; not an AWS post-incident report.","architectureImplications":"Dependency of AI services on regional compute/storage control planes; supports multi-Region inference profiles and graceful degradation for AI front ends.","governanceImplications":"Agencies should define AI service availability expectations and fallback procedures in runbooks and contracts.","securityPrivacyImplications":"Availability risk only; no security impact reported.","caveats":"Which Bedrock functions degraded is not specified; probe methodology not detailed; times differ between official window (2:35–4:21 PM PST) and ModelBank window; no AWS root-cause report.","streamIds":["aws"],"roles":{"sales":"Raise resilience proactively with public-sector AI buyers: a November 6 Oregon EC2/EBS incident degraded Bedrock for a period, per independent monitoring. Discovery should ask which Region hosts their Bedrock workloads, whether cross-Region inference is enabled, and what residents see when AI is unavailable. Avoid overstating the outage or claiming GovCloud impact. A bounded engagement is a resilience review of a constituent-facing AI service with a documented fallback and Region strategy.","engineering":"Fit: any SLED Bedrock workload pinned to US-West-2. Architect for regional events by using cross-Region inference profiles where data rules allow, timeouts and circuit breakers in AI front ends, and non-AI fallbacks such as static FAQs or human queues. Prerequisites are Region inventory and data-residency rules. Constraints: cross-Region routing may conflict with residency requirements; GovCloud has its own cross-Region option. Proof of value is a game-day test showing graceful degradation when Bedrock calls fail or slow.","delivery":"Add AI-specific incident playbooks. Dependencies include CloudWatch alarms on Bedrock invocation errors and latency, AWS Health notifications and a fallback content path. Ownership sits with the service operations team. Governance checkpoints: approve cross-Region routing against data rules; review fallback messaging for accessibility. Acceptance criteria, as proposals: alarms fire within five minutes of elevated errors; fallback activates in game-day testing. Risks are untested fallbacks and residency conflicts from cross-Region routing."},"roleDetails":{"sales":{"customerProblem":"Constituent-facing AI services can degrade when underlying regional AWS infrastructure has incidents.","stakeholders":["Agency CIO","Service operations lead","Constituent services director"],"discoveryQuestions":["Which Region hosts your Bedrock workloads?","Is cross-Region inference allowed under data rules?","What is the non-AI fallback?"],"valueHypothesis":"Resilience design can limit resident-facing disruption during regional incidents; not quantified here.","potentialEngagement":"Resilience review and game-day test for one AI service.","unsupportedClaims":["GovCloud impact","Specific Bedrock features affected","Root cause beyond AWS status text"]},"engineering":{"fit":"Relevant to US-West-2 Bedrock workloads; general resilience lesson elsewhere.","architectureIntegration":"Cross-Region inference profiles, timeouts, circuit breakers, fallback paths.","prerequisites":["Region inventory","Data-residency rules","CloudWatch alarms"],"constraints":["Residency limits on cross-Region routing","Third-party monitoring data"],"security":"Availability only; no security impact reported.","proofOfValue":["Game-day shows graceful degradation","Alarm latency measured"]},"delivery":{"work":"Create AI incident playbook and fallback path.","dependencies":["CloudWatch","AWS Health","Fallback content"],"ownership":"Service operations team.","skillsAdoption":"Incident response for AI services.","governanceCheckpoints":["Cross-Region routing approval","Fallback messaging review"],"acceptanceCriteria":["Proposed: alarms within five minutes","Proposed: fallback activates in game-day"],"risks":["Untested fallback","Residency conflict"]}},"retrievedAt":"2026-11-08T04:05:51Z","enrichedAt":"2026-11-08T04:11:17Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Constituent-facing AI assistants should have non-AI fallback paths during degradation.","procurementImplications":"Consider availability commitments and incident communication terms when procuring AI services.","operatingModelImplications":"Operations teams need monitoring of Bedrock-dependent workflows and incident playbooks.","sourceVerification":{"openedUrl":"https://modelbank.ai/outages/bedrock-2026-11-06","referenceExcerpt":"On November 6, 2026, Amazon Bedrock had degraded performance between 22:51 and 01:12 UTC, about 2.4 h in total.","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}