Cases26 cases

4 to 8 of 26

1. IL-0413 LLM inference engines timed out or ran out of resources under load in about 60% of 156 severe incidents at Microsoft

   Microsoft2025

   Symptoms

   In Microsoft's cloud LLM inference service, the inference engines themselves failed in 94 of the 156 high-severity incidents from April to June 2025 (≈60%). Requests timed out or ran into resource exhaustion under load, and callers got HTTP 408, 429 and 5xx errors.

   Intermittent timeoutsHigh error rateHigh latency

   Trigger

   Bursty load, traffic with unusually high sampling parameters and long-running requests pushed the inference engines into resource exhaustion and timeouts.

   Root cause

   Capacity exhaustion

   Component

   Model inference server

   Contributing factor

   Monitoring blind spot

   Detection

   Across the 156 incidents, 115 (≈74%) were auto-detected, meaning health probes or DRI pipelines alerted before customers reported the problem; the others were not.

   Remediation

   • Enabled dynamic pushback at the load balancer with caps on pending tokens and requests
   • Tuned request batching and GPU/CPU replica counts
   • Improved gRPC backoff and error mapping
   • Limited sampling parameters and throttled per client
   • Enabled streaming, tuned request batching and capped pending tokens against engine connection timeouts on long-running requests
   • Increased capacity, routed traffic to healthy endpoints and restarted nodes, the actions applied mostly to inference engine failures

   Action items

   • Optimise serving for long-running requests and high sampling regimes, with connection liveness and GPU capacity-aware routing as the main levers
   • Expand health-probe-driven failover and GPU capacity-aware routing so traffic moves off degraded endpoints on its own
   • Forecast token and throughput growth and automate capacity increase decisions
   • Deploy adaptive alerting and expand engine-level telemetry to close detection gaps

   Lessons

   Hotfixes restore correctness but lengthen time to mitigate, while traffic routing and node rebalancing mitigate faster under overload and timeout patterns. Even with AIOps tooling a meaningful share of incidents still needs manual detection or judgement, so health-probe-driven traffic routing, restarts and capacity increases are the decisions to automate first.

   View case
2. IL-0048 DNS lookups timed out in a large EKS cluster when CoreDNS hit the 1024 packets-per-second link-local limit, driven by sudo hostname lookups

   unknown2025

   Symptoms

   CI jobs failed now and then because a hostname could not be resolved. The failures came from tensorstore, appeared on several different instances and did not repeat consistently. The error read: "./tensorstore/util/result.h:420: Status not ok: status(): UNAVAILABLE: CURL error Could not resolve hostname: Could not resolve host: company-bucket.s3.amazonaws.com".

   DNS resolution failuresIntermittent timeouts

   Trigger

   An administrative pod running with host networking called sudo in a loop. Each sudo call looked up the node's own hostname, and with ndots:5 and four search domains every lookup became five queries, for both A and AAAA records.

   Root cause

   Capacity exhaustion

   Component

   CoreDNSClient-side DNS resolver and cacheVPC DNS resolver

   Contributing factor

   Unknown limitUnreviewed defaults

   Detection

   Reports of flaky CI jobs that could not resolve an S3 bucket hostname, followed by i/o timeout errors in the CoreDNS logs.

   Remediation

   • added 'Defaults !fqdn' in /etc/sudoers.d/00-no-fqdn (mode 440) to the container image so sudo stopped looking up the hostname

   Lessons

   Traffic from CoreDNS to the VPC DNS resolver is limited to 1024 packets per second per interface, and drops show up only as UDP timeouts; the linklocal_allowance_exceeded counter in ethtool shows them. Names that are not fully qualified are multiplied by the search domains under ndots:5, and AAAA lookups double them again. sudo looks up the host's own name on every call unless 'Defaults !fqdn' is set.

   View case
3. IL-0247 One in a million search requests took 100 times longer on Pinterest's Manas leaf nodes after moving to Kubernetes

   Pinterest2025

   Symptoms

   During performance testing of Pinterest's search system Manas on its new Kubernetes platform, a small but persistent share of search requests timed out. About one request in a million took 100 times longer than usual.

   High latencyIntermittent timeouts

   Root cause

   Software defect

   Component

   Telemetry agentLinux kernel

   Contributing factor

   Unreviewed defaults

   Detection

   Performance testing with traffic sent to a test cluster showed a small but persistent percentage of requests timing out.

   Remediation

   • Disabled cAdvisor's working set size estimation on all PinCompute nodes

   Lessons

   CPU shielding does not fully isolate a process from its neighbours. Showing early that the node image was not the cause narrowed the search and avoided dead ends. Simple blackbox debugging, breaking things until the system stops working, works as a way of bisecting the problem.

   View case
4. IL-0080 kube-dns on Mercari's GKE clusters could not keep up with peak-hour DNS query rates, and a Sev1 production outage followed when several services failed DNS resolution

   Mercari2025sev1

   Symptoms

   During peak hours, when Mercari's GKE clusters handle very high request rates, the platform team saw kube-dns reaching its limits. DNS query rates were high enough to overwhelm the kube-dns service.

   DNS resolution failuresIntermittent timeoutsService outage

   Trigger

   Peak-hour traffic on the GKE clusters produced more DNS queries than the few kube-dns pods could answer.

   Root cause

   Capacity exhaustion

   Component

   kube-dns

   Contributing factor

   No cachingSingle point of failure

   Detection

   During peak hours the team saw frequent NXDOMAIN and truncated-response errors and recurring "failed to refresh DNS cache" errors on the clusters, which showed that kube-dns had reached its limits.

   Remediation

   • deployed NodeLocal DNSCache on the Laboratory GKE cluster with a specific nodeAffinity to measure its effect
   • rolled NodeLocal DNSCache out gradually to the Dev and Prod environments by labelling node pools so its pods could be deployed there

   Lessons

   When every pod in a busy cluster resolves names through a few central kube-dns pods, DNS becomes a shared bottleneck, and a DNS failure quickly turns into higher latency, timeouts and cascading outages across microservices. A cache on each node takes most of that load off kube-dns and keeps resolution local to the node. Rolling the cache out first to an internal test cluster with nodeAffinity, and then node pool by node pool through labels, kept the blast radius of the change small.

   View case
5. IL-0201 A framework upgrade caused kernel panics in GitHub's Kubernetes infrastructure, and GitHub.com web and API requests, mostly for issues and pull requests, failed intermittently for about 16 hours

   GitHub2025

   Symptoms

   From 18:05 UTC on March 17, 2025, GitHub.com web and API requests failed now and then for a small percentage of users, mostly requests about issues and pull requests. At 18:39 UTC the status page reported degraded performance for Issues.

   High error rateIntermittent timeouts

   Trigger

   A framework upgrade rolled out to GitHub's Kubernetes infrastructure caused kernel panics.

   Root cause

   Breaking change in an upgrade

   Component

   Linux kernel

   Contributing factor

   –

   Detection

   The status page opened the incident at 18:39 UTC, 34 minutes after the failures began, as an investigation of reports of degraded performance for Issues.

   Remediation

   • Downgraded the framework upgrade that caused the kernel panics
   • Disabled the problematic feature of the framework

   Action items

   • Investigate why the upgrade caused the kernel panics
   • Take steps to temporarily prevent the problem from coming back
   • Work out longer-term patch plans
   • Improve observability so that similar classes of problems can be reacted to quickly

   View case

By root cause7