Cases43 cases

1 to 5 of 43

1. IL-0136 A deployment that cleared an internal service's caches at peak traffic sent 20 times the normal load to the Azure OpenAI metadata backend in Sweden Central, and health checks that depended on its database restarted instances in a loop for almost six hours

   Microsoft Azure2026

   Symptoms

   From 10:03 UTC on 29 September 2026, requests to Azure OpenAI Service, Foundry Agent Service, Foundry Models and Azure AI services in Sweden Central failed intermittently, took longer, or returned HTTP 500 and 503 errors. At 10:17 UTC service monitoring raised an alert on elevated failure rates for Azure OpenAI Service requests in the region.

   High error rateHigh latencyIntermittent connection failuresIntermittent timeoutsProbe failuresOutbound connection exhaustion

   Trigger

   At about 09:38 UTC a planned deployment of an internal service in Sweden Central cleared that service's caches during the region's peak traffic hours.

   Root cause

   Faulty deployment

   Component

   In-memory cacheApplication database

   Contributing factor

   Change during peak hoursRetry stormAutomation amplificationMonitoring blind spotNo runbookHealth check tied to a dependency

   Detection

   Service monitoring detected elevated failure rates for Azure OpenAI Service requests in the region at 10:17 UTC and alerted the engineering teams; engineers were engaged at 10:28 UTC.

   Remediation

   • reconfigured the backend service's automated health check so it no longer depended on remote connectivity, which stopped the restart loop

   Action items

   • Reconfigure the backend service's automated health check so it no longer depends on connectivity to its database layer
   • Add monitoring and alerting on the load and latency of the caching layer used by the backend service
   • Update troubleshooting and procedural guidance for connection exhaustion scenarios for on-call engineers
   • Add throttling and rate limiting to protect the backend service from sudden traffic increases (estimated October 2026)
   • Increase the resource capacity of the caching layer used by the backend service across regions (estimated October 2026)
   • Improve diagnostic metrics and automated detection to identify such issues more quickly (estimated October 2026)
   • Partition the caching layer used by the backend service to increase its throughput (estimated December 2026)

   Lessons

   A health check that depends on a remote dependency turns a dependency problem into a restart loop: instances that cannot reach the database are killed, the remaining ones take more load and retries, and adding instances or resources does not help until the check itself is changed. Clearing a cache during peak traffic can multiply load on the backend behind it, especially when cache-read defects let lookups bypass the cache. Connection exhaustion was hard to see because the existing network and diagnostic metrics did not show it.

   View case
2. IL-0071 Worker node running out of inodes while disk space looked fine, because containerd kept a full node_modules tree in each of several image layer snapshots

   Onkar Raghunath Shelke2026

   Symptoms

   A Kubernetes worker node paged with NodeFilesystemFilesFillingUp, the node_exporter mixin alert that predicts a filesystem will run out of inodes. The engineer checked disk space first: df -h showed the root filesystem 83% used, 103G used and 22G available. df -i showed inodes 67% used, 11M used and 5.3M free.

   Node disk full

   Trigger

   Several versions of a single-stage Node.js image, each shipping more than 40,000 files, were held on the node; containerd unpacked every layer with a distinct digest into its own snapshot, each with a full copy of node_modules.

   Root cause

   Capacity exhaustion

   Component

   Node root diskkubeletContainer runtime

   Contributing factor

   Unreviewed defaultsMonitoring blind spotUnknown limit

   Detection

   The NodeFilesystemFilesFillingUp alert from the node_exporter mixin fired on the predicted inode trend while inode use was at 67%.

   Action items

   • Build the image in multiple stages so the final image contains only the compiled assets
   • Add a flat alert on root filesystem inode use above 80% for 30 minutes alongside the predictive alert
   • Put a Prometheus query on a dashboard that shows, per node, inode usage minus disk usage, sorted from the highest
   • Remove unused images from nodes with crictl rmi --prune to free old snapshots

   Lessons

   kubelet's background cleanup works on bytes only. A workload heavy in small files gets no early intervention for inodes, and the first thing kubelet does at 5% free inodes is evict pods. Watch inode use on nodes separately from disk use, and keep dependency trees out of runtime images.

   View case
3. IL-0102 CoreDNS pods on a production EKS cluster ran out of memory under a burst of new pods, endpoints and nodes, taking down the Buildkite web UI, APIs and job dispatch

   Buildkite2026

   Symptoms

   From 22:47 UTC on 25 August 2026 the Buildkite web interface, REST API, Agent API, job queue and notifications stopped working for every customer. No one could create a build from the UI or the REST API, job dispatch was delayed, and some builds that were already running when the outage began stayed stuck until the services recovered.

   Service outageDNS resolution failuresOOMKilledAutoscaler misbehaviour

   Trigger

   At 22:44 UTC an application deploy created a surge of new pods, which used up the free capacity on the existing nodes and made the EKS cluster scale out. Background worker pools with unusually high maximum replica counts were scaling out at the same time and kept asking for more pods while new nodes were still being provisioned. Together they produced an unusually high rate of change in applications, network endpoints and nodes, and with it a sudden rise in demand on CoreDNS.

   Root cause

   Capacity exhaustion

   Component

   CoreDNSContainer memory limitHorizontal pod autoscalerNode pool

   Contributing factor

   No headroomAutomation amplificationRetry stormMonitoring blind spot

   Detection

   The first status post, at 22:55 UTC, eight minutes after the outage began, said something had gone wrong, and at 23:05 UTC the status page said a sitewide outage was being investigated. CoreDNS query duration had been rising as the cluster grew, but a defect in a key monitoring query and too little monitoring of query volume meant this went unnoticed before the outage.

   Remediation

   • paused further application deploys
   • scaled CoreDNS from three to twelve pods
   • raised each CoreDNS pod's memory limit from 512 MiB to 2 GiB
   • expanded the system node pool that runs CoreDNS

   Action items

   • Increase immediate CoreDNS headroom: twelve pods instead of three, a 2 GiB memory limit instead of 512 MiB, and a larger system node pool
   • Enable EKS-managed CoreDNS autoscaling with a tested minimum replica count and enough capacity to spread the replicas across nodes and availability zones
   • Add direct alerts on CoreDNS query duration, query volume, goroutine growth, memory pressure, OOM restarts and available replicas, and track longer-term trends in capacity planning
   • Find other cluster-critical services that lack tested capacity, direct alerting, spread across failure domains, and either safe autoscaling or documented static headroom, and fix them

   Lessons

   A CoreDNS deployment sized once and left at a fixed size can carry a growing cluster for months and then fail on the first burst of churn it cannot absorb. When query processing slows, pending queries take up memory, the pods hit their limits and restart, and client retries keep the service down after the restart. Server-side DNS query duration and volume need their own alerts: here client-side latency was masked by unrelated application speed-ups, and a broken monitoring query hid the server-side trend. Maximum replica counts raised as a temporary measure remove the cap that normally stops runaway scaling when new nodes are slow to arrive.

   View case
4. IL-0133 A storm shut down cooling at Nebius's us-central1 data center, and a disk hot-plug bug kept Managed Kubernetes control planes down for hours after power returned

   Nebius2026

   Symptoms

   On 19 August 2026 the first equipment-level signs in Nebius's us-central1 region appeared around 09:40 UTC: the InfiniBand fabric degraded and GPU over-temperature alerts began to accumulate. Customer reports followed shortly after 10:00. At 10:15 on-call engineers were paged on critical component temperature alerts, and the incident was declared at 10:21.

   Control plane unresponsiveService outageMonitoring outage

   Trigger

   A storm-related event at the data center building took the building management system offline and shut down the chilled-water cooling loop.

   Root cause

   Dependency failure

   Component

   Data center power and coolingCloud block storageetcdInfiniBand fabric

   Contributing factor

   Single point of failureMonitoring blind spotNo alert on limitNo runbookOverlapping failures

   Detection

   No facility-level alert fired, because the storm also disabled the building management system that monitors the facility. Engineers acted on component over-temperature alerts that paged on-call at 10:15 UTC, after InfiniBand degradation, GPU over-temperature alerts and customer reports; the temperature rise from about 09:15 was visible on dashboards but had no alert.

   Remediation

   • had the facility operator switch cooling to manual bypass mode and opened rack rear doors to vent heat
   • re-energised tripped racks, control-plane racks first
   • restored the regional infrastructure control plane by manual intervention
   • restarted block-storage host components to clear stale disk attachment state
   • raised the VM recovery rate limits by hand for the region-scale recovery
   • started the affected Managed Kubernetes control-plane instances by hand
   • applied a controlled stop and start to control-plane instances whose data disk had been left unattached
   • identified and started 104 customer VMs that the incident had stopped

   Action items

   • Complete a joint resilience review with the facility operator covering redundancy of facility-management and monitoring systems, cooling that keeps running when those fail, electrical protection of cooling equipment and on-site response
   • Add facility-level environmental alerting on cooling plant state and room temperature trend that pages on-call independently of the building management system
   • Set up a documented emergency escalation path with the facility operator, with an on-call contact tree and guaranteed response times
   • Define an emergency heat-load reduction procedure with criteria, ownership and mechanics for shutting down or degrading workloads during a cooling loss
   • Establish a dependency-ordered region cold-start procedure with clear ownership for each service layer
   • Run regular region-scale recovery drills that cover VMs, storage attachments and Managed Kubernetes clusters
   • Fix the disk-attachment bug that extended the Managed Kubernetes outage and deploy the fix to all production regions
   • Improve the recovery policies Compute users can configure, and move all Managed Kubernetes control-plane nodes to the improved policy
   • Automate reconciliation of stale storage attachment state after abrupt host failures
   • Remove in-region dependencies from monitoring startup and recovery, and extend cross-region monitoring
   • Make regional health checks used for global traffic management reflect real service availability

   Lessons

   Facility monitoring that runs on the same system it watches goes silent in exactly the failure it should report, and with no alert on the room temperature trend there was no practical window to shed heat load in a controlled way. A cold start of a whole region exercised recovery paths built for isolated failures: single-attempt VM recovery, conservative rate limits and stale disk attachments all needed manual work. The longest part of the outage was restoring customer resources, Managed Kubernetes control planes above all, so Nebius is extending its recovery drills to that layer.

   View case
5. IL-0023 A training run crashed with an NCCL timeout after one InfiniBand port went down once and recovered

   Databricks2026

   Symptoms

   Seven hours into a distributed training run on Databricks AI, the job failed with an NCCL timeout.

   Training job failureNCCL timeoutLink flapping

   Trigger

   A single InfiniBand port used for RDMA NCCL collectives went down once and recovered.

   Root cause

   Dependency failure

   Component

   NCCLInfiniBand fabricPyTorch

   Contributing factor

   Monitoring blind spotUnreviewed defaults

   Detection

   The training job crashed with the NCCL timeout. The continuous health checks, which watched InfiniBand port flapping by count, had not flagged the single flap.

   Remediation

   • Tuned the NCCL_IB_TIMEOUT defaults to be more resilient to short port-down windows
   • Used the port-down signal to crash the job and restart it from a checkpoint instead of waiting for the watchdog
   • Fed the investigation into gpu-monitor, whose RDMA fabric port-down check is thresholded on cumulative downtime rather than flap count

   Action items

   • Tune the NCCL_IB_TIMEOUT defaults to be more resilient
   • Crash and restart the job from a checkpoint on the port-down signal instead of leaving GPUs idle until the watchdog fires
   • Threshold RDMA fabric port-down detection on cumulative downtime, not flap count, in gpu-monitor

   Lessons

   For training impact the signal that matters is cumulative port downtime, not the number of flaps: one long enough flap can crash a multi-day run. NCCL_IB_TIMEOUT, with an effective default of about seven seconds, decides the outcome long before the PyTorch watchdog timeout does. Health checks need to cover the whole node lifecycle: active checks before workloads start, passive checks for silent degradation under load, and periodic multi-node NCCL fabric checks in between.

   View case

By root cause9