Cases20 cases

11 to 15 of 20

1. IL-0176 Prometheus in a 235-node Kubernetes cluster crash-looped because over two million series exhausted host memory and the kernel OOM killer chose it first, as Burstable QoS gave it oom_score_adj 999

   LINE2020

   Symptoms

   One Monday morning the author opened the Prometheus dashboard to check a Kubernetes cluster for alerts and could not reach it. The Prometheus pod was in CrashLoopBackOff: kubectl get pods showed it 1/2 ready with 950 restarts over 3d20h on worker3.

   Container crash loopOOMKilledMonitoring outageProbe failures

   Trigger

   After start, the Prometheus time series count rose to 2,162,256 within five minutes and its memory use filled the 8GB host.

   Root cause

   Capacity exhaustion

   Component

   PrometheuskubeletLinux kernel

   Contributing factor

   No resource requests or limitsNo capacity planning

   Detection

   The author found it on a Monday morning when the Prometheus dashboard, opened to check the cluster for alerts, could not be reached.

   Remediation

   • removed most of the unnecessary settings from the Prometheus ConfigMap, after which the container stopped crashing

   Action items

   • Set appropriate resource requests and limits on the Prometheus pod
   • Reduce unneeded metrics by adjusting scrape targets or scrape interval
   • Estimate Prometheus memory needs, provide dedicated hosts with a node label such as deploy-topology: Prometheus, and pin the Prometheus pod to them with a nodeSelector

   Lessons

   Exit code 137 together with an OOM event in the kubelet log means the kernel killed the container, not kubelet. Under node memory pressure the kernel kills the process with the highest score, and kubelet derives each container's oom_score_adj from the pod's QoS class: a container without a memory request in a pod where another container has one ends up Burstable with a score near 1000, so it is killed first. In a large cluster, Prometheus memory needs to be estimated from the expected series count and requested explicitly.

   View case
2. IL-0012 CPU limits and an unpatched kernel CFS quota bug throttled multi-threaded services into stalls, failed health checks and timeouts

   Omio2020

   Symptoms

   For about six months, applications on Omio's Google Kubernetes Engine clusters stalled at random: they got stuck, stopped answering health check requests or lost network connections, and nothing the team looked at explained why.

   CPU throttlingProbe failuresHigh error rateHigh latencyIntermittent timeoutsIntermittent connection failures

   Trigger

   Multi-threaded workloads running with CPU limits on nodes whose kernel did not yet have the fix for a known CFS quota bug that causes unnecessary throttling.

   Root cause

   Software defect

   Component

   Container CPU limitLinux kernelkubelet

   Contributing factor

   Unknown limitMonitoring blind spotUnpatched known bug

   Detection

   Months of random stalls and failed health checks led the team to read the cgroup cpu.stat throttling counters inside the pods, which showed far more throttling than the limits should allow.

   Remediation

   • removed CPU limits until the kernel with the CFS bug fix was deployed across all clusters

   Action items

   • Deploy the kernel with the CFS quota bug fix across all clusters
   • Monitor utilisation, saturation and error metrics, API latency and error rates to confirm the change behaves as expected

   Lessons

   A CPU limit is a time quota per 100 ms period, not a share of cores, so a multi-threaded container can be stopped for most of each period long before it uses the cores it was given, and a kernel bug can make that worse. Throttling is invisible in standard Kubernetes metrics and has to be read from the cgroup counters. Removing limits trades away protection against a container that takes a whole node, which is only acceptable when runtimes are tuned.

   View case
3. IL-0008 An unresponsive OPA validating webhook crash-looped the API server during a GKE master upgrade, and node auto-repair then recreated every node

   Jetstack2019

   Symptoms

   Jetstack was upgrading the master of a development cluster that many teams use to test their apps during the working day, a regional Google Kubernetes Engine (GKE) cluster in europe-west1. The upgrade ran through Jetstack's GKE Terraform pipeline and did not finish before the 20-minute Terraform timeout, while the GKE console still showed the cluster as upgrading.

   Upgrade stallControl plane unresponsiveNode recreation loopService outage

   Trigger

   A planned master upgrade of the shared development cluster. The same upgrade had been done on another pre-production environment earlier that day, and no API downtime was expected.

   Root cause

   Configuration change

   Component

   Admission webhookGKE control planeKubernetes API serverkubelet

   Contributing factor

   Excessive blast radiusNo health probesMonitoring blind spotAutomation amplification

   Detection

   The first sign was the master upgrade missing the 20-minute Terraform timeout; re-running the pipeline returned the kube-apiserver unhealthy error. The cause was identified with the help of Google Support.

   Remediation

   • Deleted the OPA ValidatingAdmissionWebhook resource during a window of API server access

   Action items

   • Limit the OPA webhook to the namespaces where the policy applies and development teams have access
   • Enable the webhook only for Ingress and Service resources, the only resources the policy validates
   • Add a liveness probe so OPA is restarted when it stops responding, and update the OPA documentation
   • Follow the open upstream Kubernetes issue about this behaviour

   Lessons

   With alerting on API server response times, the team might have seen CREATE and UPDATE requests slow down across the board as soon as the OPA-backed webhook was first deployed. Every workload needs probes: OPA looked so simple to deploy that the team skipped the Helm chart, which adds a livenessProbe for the admission controller containers among other adjustments to the tutorial configuration. Others had hit the same problem, and an upstream issue to improve it was still open.

   View case
4. IL-0002 CI builder pods could not get AWS IAM credentials because kubelet fell minutes behind its API rate limit

   Zalando2019

   Symptoms

   Builds on Zalando's in-house Continuous Delivery Platform (CDP), which runs CI/CD on Kubernetes, timed out and failed for almost an hour on 11 June 2019. The Kubernetes builder pods could not get AWS IAM credentials.

   High error rateCredentials unavailableControl plane lag

   Trigger

   An unlucky rescaling of the CDP Kubernetes cluster left only one node available for builder pods. Builder pods were created and deleted on that node so quickly that kubelet fell behind.

   Root cause

   Capacity exhaustion

   Component

   kubeletCredential broker for podsKubernetes API serverCI system

   Contributing factor

   Unreviewed defaultsSingle point of failure

   Remediation

   • Scaled the CDP cluster up by hand so builder pods spread over more than one node

   Action items

   • Raise the kubelet QPS limit from the default 5 to 50 in the global cluster configuration

   View case
5. IL-0035 Worker nodes went NotReady after a new ElastAlert rule without memory limits used up node memory

   NU.nl2019

   Symptoms

   The NU.nl development team got reports that several services on one of its Kubernetes clusters behaved erratically: occasional error pages, slow responses and timeouts. The affected services were mostly internal tools and dashboards.

   Node NotReadyHigh error rateHigh latencyIntermittent timeoutsMonitoring outageOOMKilled

   Trigger

   A new ElastAlert rule, deployed shortly before from the command line, that fired if the Cloudfront-* Elasticsearch indexes stopped receiving documents; those indexes take a few million requests per hour, and the rule made ElastAlert's memory grow without bound.

   Root cause

   Configuration change

   Component

   Container memory limitkubeletPrometheus

   Contributing factor

   No resource requests or limitsUnreviewed defaultsNo alert on limitMonitoring blind spotExcessive blast radius

   Detection

   Users reported error pages, slow responses and timeouts from several services; no alert fired.

   Remediation

   • terminated the NotReady nodes one by one so the auto-scaling group replaced them (the new nodes also went NotReady, but Prometheus and Grafana landed on a stable node)
   • found the ElastAlert pods' memory growth on the Grafana pod memory dashboard
   • fixed the ElastAlert problem that caused the memory growth

   Action items

   • Define alerts on resource usage such as CPU, memory and disk space
   • Keep researching an alerting strategy that combines Elasticsearch, Prometheus Alertmanager and New Relic
   • Consider exporting metrics out of the cluster and running Grafana outside it, on EC2 or a separate Kubernetes cluster
   • Either visualise the important metrics in the ELK stack or make Prometheus and Grafana more available, and improve the metrics strategy
   • Set resource limits for ElastAlert through the Helm values
   • Configure namespaces with default memory requests and limits
   • Isolate customer-facing workloads from build, logging and monitoring workloads (a dedicated node group, or later a separate tools cluster)
   • Deploy everything through automation such as Jenkins pipelines, and consider a GitOps approach
   • Deploy through a pipeline that includes a smoke test step
   • Make Kubernetes and cloud infrastructure work part of the team's sprints, picked up by all team members while pairing with more experienced members, and hold workshops that go deep into specific topics

   Lessons

   Fixing the root cause covers only a small part of what an incident can teach; the chain of smaller failures around it is where most of the learning is. A pod without a memory limit can take whole nodes down, so limits need defaults at the namespace level. Monitoring that runs inside the cluster it watches goes blind with it. Changes installed by hand from the command line leave the rest of the team without a record of what changed.

   View case

By root cause7