Cases14 cases

11 to 14 of 14

1. IL-0199 A non-essential container exceeding its memory limit made GitHub's production pods unavailable, and their restarts failed because a DNS maintenance change had left the clusters unable to reach the image registry

   GitHub2020

   Symptoms

   At 08:18 UTC on 13 July 2020, production Kubernetes pods began to be marked unavailable. The loss of pods spread through the clusters and reduced capacity until GitHub.com services went down.

   OOMKilledImage pull failuresService outage

   Trigger

   A container in the production pods exceeded its memory limit and was terminated; a routine DNS maintenance operation completed earlier had left the clusters unable to reach the image registry.

   Root cause

   DNS or network fault

   Component

   Container memory limitContainer registry and its credentials

   Contributing factor

   Overlapping failuresHarmful manual actionSingle point of failure

   Remediation

   • restarted the process with the cached DNS records so the clusters could reach the registry
   • fetched the container images, redeployed and recovered the services

   Action items

   • Improve monitoring so pod restarts cannot fail again in the same pattern
   • Minimise the dependency on the image registry
   • Expand validation during DNS changes
   • Re-evaluate all existing Kubernetes deployment policies
   • Improve progressive deployments so the impact of a deployment can be evaluated incrementally

   Lessons

   A container that production traffic does not need can still take pods out of service, because a pod is available only when all of its containers are healthy. ImagePullPolicy Always turns every container restart into a registry dependency, so an unrelated DNS change that breaks registry access turns a one-minute self-healing event into a long outage, and a mass redeploy spreads it.

   View case
2. IL-0193 A GCP persistent disk degradation took down a Cassandra StatefulSet whose OrderedReady policy blocked recovery, and Metrictank could not reconnect to the recreated pods, causing a 23-hour Graphite query outage

   Grafana Labs2019

   Symptoms

   An alert fired because some Metrictank instances could not drain their write queue, which meant they could not write to Cassandra. Google had not yet published an incident, but a team member's Discord server, also hosted on GCP, was down, which pointed at GCP; Google confirmed an SSD persistent disk incident about 90 minutes after it began.

   High error rateNode NotReadyPods PendingService outage

   Trigger

   A Google Cloud incident in which SSD persistent disks performed so poorly that many I/O operations failed for an extended period.

   Root cause

   Cloud provider outage

   Component

   Persistent volumes and stateful setsApplication databaseNode root disk

   Contributing factor

   Overlapping failuresUnpatched known bugUnverified assumption

   Detection

   An alert that Metrictank write instances could not flush their write queue to Cassandra. Error rates alone rose too slowly to notice quickly, because most queries were served from memory.

   Remediation

   • Added compute nodes hoping new nodes could attach the disks, which did not work
   • Deleted and recreated the Cassandra StatefulSet with an updated pod management policy, so all pods but one started
   • Raised Kafka topic retention to avoid data loss during the long downtime
   • Temporarily added nodes so Metrictank restarts were not slowed by CPU starvation
   • Restarted all Metrictank instances, which took more than 20 hours, so they connected to the new Cassandra IPs

   Action items

   • Re-instantiate the gocql connection in Metrictank when it is detected dead, so it reconnects even when all Cassandra IPs change
   • Fix the Metrictank write-queue deadlock by updating an internal data structure with atomics instead of a lock

   Lessons

   An OrderedReady StatefulSet cannot recover past a single pod whose volume will not attach; when every pod is down, the ordering that protects normal start-up blocks the whole cluster. A client that learns peer addresses only from live peers needs a way to go back to the service name when all peers are replaced at once. A long-tolerated bug with a config workaround can surface at the worst time; this one turned out to have a simple fix. Writing incoming data to a Kafka write-ahead log first meant all data could still be ingested after recovery, and Kafka on non-SSD disks escaped the disk incident. With both Metrictank bugs fixed, the team expects a similar event to take less than 3 hours to recover from instead of 23.

   View case
3. IL-0014 A GKE upgrade to a withdrawn version stalled on slow drains and a stuck node pool, then broke an ingress and confused the HPA

   loveholidays2019

   Symptoms

   On the evening of a long-overdue upgrade of a loveholidays production GKE cluster from 1.10.X to 1.12.5-gke.10, planned to take 2 to 2.5 hours, the master upgrade and the first two 7-node NodePools went smoothly. After that the NodePool upgrades slowed to a crawl. Each node is cordoned, drained and upgraded in turn. Even a 14-node cluster at about 5 minutes per node took over an hour, and with about 100 more nodes to go, some of them taking over 15 minutes to drain, the sequential drains were too slow to finish at a reasonable hour.

   Upgrade stallPods PendingScale-up failureService outageAutoscaler misbehaviour

   Trigger

   A manual upgrade of the GKE masters and NodePools from 1.10.X to 1.12.5-gke.10, two versions at once, on 19 March 2019; GKE's release notes of 14 March had already said that version was no longer available for new clusters or master upgrades because masters showed elevated error rates, but it could still be selected.

   Root cause

   Cloud provider outage

   Component

   GKE control planeNode poolCluster autoscalerIngress controllerHorizontal pod autoscaler

   Contributing factor

   Unrepresentative stagingOverlapping failuresIgnored early signal

   Detection

   The slow drains, zombie nodes and stuck NodePool were seen by the engineer running the upgrade; the broken Ingress was noticed after the upgrade and explained by Google support; the HPA problem was noticed the next day.

   Remediation

   • drained the remaining NodePool nodes ahead of the upgrade with kubectl drain --ignore-daemonsets --delete-local-data
   • deleted the zombie instances from the managed instance group with gcloud compute instance-groups managed delete-instances
   • cancelled the stuck UPGRADE_NODES operation with gcloud beta container operations cancel
   • used kubectl port-forward in place of the broken internal Ingresses
   • scaled down the older of two active ReplicaSets, or triggered a dummy deployment, to bring the HPA back

   Action items

   • Alert when a HorizontalPodAutoscaler is at 70% and at 100% of its capacity
   • Alert when a deployment has had two active ReplicaSets for 30 minutes and for 60 minutes
   • Re-evaluate termination grace periods that were set to arbitrary large values
   • Read the GKE release notes before every cluster upgrade
   • Use auto-upgrade in non-production environments, and manual cluster upgrades with automatic node upgrades in production
   • Keep environments as identical as possible, including resource names

   Lessons

   Read the release notes before upgrading: the target version had been withdrawn five days earlier. Drain time depends on pod configuration, so emptyDir volumes, long grace periods and missing disruption budgets turn a NodePool upgrade into hours. Pre-production only catches problems like the ingress bug if it matches production down to resource names. Auto-upgrades in non-production surface problems early, and a mix of manual cluster upgrades and automatic node upgrades in production keeps clusters current without much effort.

   View case
4. IL-0003 linkerd lost Kubernetes service discovery after an etcd reconfiguration, and restarting every linkerd instance turned failed payments into a full platform outage

   Monzo2017

   Symptoms

   At 14:10 an engineer deployed a change to a service that Monzo's current account needs to process payments, and every request to that service started to fail. Current account customers saw payment failures; the prepaid card was not affected at first because it does not use that service. The change was rolled back at 14:12, but the errors continued.

   High error rateStale endpoint routingService outage

   Trigger

   At 14:10 an engineer deployed a change to a service needed to process current account payments. Two weeks earlier the etcd cluster had been upgraded and grown from three to nine nodes, and one day earlier another team had scaled a new service down to zero replicas while leaving its Kubernetes service in place.

   Root cause

   Software defect

   Component

   etcdKubernetes API serverLinkerdkubelet

   Contributing factor

   Ineffective rollbackHarmful manual actionOverlapping failuresMonitoring blind spotSingle point of failure

   Detection

   Payment failures began right after the 14:10 deploy. The team rolled back at 14:12 and declared an outage internally at 14:16. At 15:13, when no service could serve requests, practically all automated alerts fired.

   Remediation

   • Rolled back the payments service change (the errors continued)
   • Activated internal backup processes so most customers could still use their card
   • Restarted all linkerd instances (this made the outage worse)
   • Restarted the three Kubernetes API server processes so replacement linkerd instances could start
   • Began deploying a newer linkerd version with the fix for empty services
   • Deleted the Kubernetes service that had no endpoints, which let linkerd load service discovery again

   Action items

   • Fix the bug in Kubernetes that can trigger timeouts after a cluster reconfiguration
   • Roll out a new version of linkerd that fixes the parsing error
   • Create better health checks, dashboards and alerts for the affected components so they give clearer signals about what is wrong and prevent human error
   • Improve procedures to communicate outages internally and externally as clearly and quickly as possible

   Lessons

   A large failure in a distributed system can be very hard to understand, and well-meant human action can compound it, as restarting every linkerd instance did here by exposing a second incompatibility. Because Monzo deploys often and its network changes constantly, the deployment and request-routing systems have to be trustworthy. Monzo committed to publishing a similar post-mortem for any incident of this severity.

   View case

By root cause7