Cases5 cases

1 to 5 of 5

1. IL-0172 An AWS Tokyo data centre cooling failure took down 6 of 60 CockroachDB nodes on Kubernetes within 22 minutes, and ranges lost quorum because the Helm-deployed cluster had no availability zone locality configured

   Devsisters2021

   Symptoms

   During a team dinner, a Slack alert reported that one node of the CockroachDB cluster behind Cookie Run: Kingdom had gone down. The DevOps engineers were not worried at first, because CockroachDB is built to survive the loss of some nodes.

   Node NotReadyData lossService outage

   Trigger

   A cooling system failure in an AWS Tokyo region data centre overheated the servers in one availability zone, which held CockroachDB nodes.

   Root cause

   Cloud provider outage

   Component

   Application databaseHelm

   Contributing factor

   Unverified assumptionExcessive blast radius

   Detection

   A Slack alert reported that a database node had gone down.

   Remediation

   • copied the data outside the broken ranges into new tables with INSERT INTO ... SELECT queries that skipped the affected ranges
   • ran debug unsafe-remove-dead-replicas together with Cockroach Labs engineers, recovering one of the two lost data ranges
   • restored the last data range with the storage-layer recovery method from the previous incident

   Action items

   • Configure CockroachDB Locality so data is spread evenly across availability zones
   • Deploy with one chart and a separate values file per availability zone to set Locality
   • Replace nodes by adding Locality-configured nodes, confirming replicas in every zone, then retiring the old nodes

   Lessons

   A replication factor sized for random instance retirements does not protect against a whole availability zone failing; for a game of this scale a multi-AZ placement is required, not optional. Keeping one uniform Helm chart made zone topology hard to set, which is part of why it was left out. A repetitive chart with a values file per zone sets it reliably.

   View case
2. IL-0029 Slow user-facing services from CPU throttling on containers that used far less CPU than their limit

   Buffer2020

   Symptoms

   Services on Buffer's kops-managed cluster on AWS (about 60 nodes and 1500 containers) answered more slowly than they should have. The container throttling metric showed that many containers were being throttled whatever their CPU usage was, including containers that never came close to their CPU limit.

   CPU throttlingHigh latencyNode NotReady

   Trigger

   CPU limits on user-facing services, enforced through CFS quota by a Linux kernel with a bug that throttled containers even when they were well under their limit.

   Root cause

   Software defect

   Component

   Container CPU limitLinux kernelNode operating system image

   Contributing factor

   Unpatched known bugUnverified assumption

   Detection

   While reviewing the container throttling metric, the team noticed many containers being throttled even though their CPU usage stayed far below their limits.

   Remediation

   • removed CPU limits from all services directly or indirectly on the users' critical path
   • moved those services to dedicated tainted nodes and gave them matching tolerations
   • set each service's CPU request to its peak usage over a few months plus a margin of more than 20%, and set memory requests the same way

   Action items

   • Move nodes to a kernel that contains the CFS throttling fix (4.19 or later), which for Buffer means a kops version whose host image is Ubuntu 20.04
   • After the kernel upgrade, check whether throttling is fully gone and update the post

   Lessons

   A CPU limit can slow a service down even when the service uses a small part of it, because of how the kernel enforces CFS quota on older kernels; the throttling metric shows it, the CPU usage graph does not. Upgrading the kernel is the safer fix. Removing limits works for latency, but only together with honest CPU and memory requests, isolation of the unbounded services, and monitoring of node usage, since unbounded containers can starve the kubelet.

   View case
3. 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
4. IL-0064 CPU-heavy pods packed onto one Google Cloud node drove it to kernel panics, and the crash moved from node to node until the site was down

   Moonlight2019

   Symptoms

   On Friday, January 18, Moonlight's API began returning occasional errors that looked like connection problems with its Redis database. The API checks Redis on every authenticated request to validate the session. At the same time the Kubernetes monitoring reported some nodes and pods as unresponsive.

   High error rateIntermittent connection failuresNode NotReadyService outageSustained high node CPU

   Trigger

   The Kubernetes scheduler put several CPU-heavy pods on the same node, including all three replicas of the website and the server-side rendering pods.

   Root cause

   Capacity exhaustion

   Component

   Pod scheduling and placementLinux kernelApplication database

   Contributing factor

   Single point of failureUnverified assumptionAutomation amplification

   Detection

   API errors that looked like Redis connectivity problems, and Kubernetes monitoring that showed nodes and pods as unresponsive. Google Cloud Support later found the link between sustained 100% CPU and kernel panics.

   Remediation

   • added anti-affinity rules to all major deployments so their pods spread across nodes

   Action items

   • Keep working with Google Cloud support to find and fix the cause of the node kernel panics

   Lessons

   Three replicas give no fault tolerance if the scheduler places them all on one node; anti-affinity rules make the spread explicit. A node failure that comes from the workload itself moves with the workload when the scheduler reschedules it.

   View case
5. IL-0036 A reindex Job that kept failing exhausted one node after another until the whole cluster was down

   Universe2017

   Symptoms

   Shortly after a search reindex job was started, problems with the Kubernetes cluster hit Universe's reporting and search APIs and the Kubernetes cluster UI itself. The status page opened an investigation into reports of search and reporting being unavailable at 14:37 EDT on 2017-07-21.

   Service outageNode NotReadyNode recreation loop

   Trigger

   A change to the search index definition, which removed some fields from full-text indexing, was shipped, and a reindex job was started as a Kubernetes Job to apply it.

   Root cause

   Capacity exhaustion

   Component

   Jobs and CronJobsPod scheduling and placement

   Contributing factor

   Unverified assumptionAutomation amplificationExcessive blast radius

   Detection

   The team saw problems with the Kubernetes cluster shortly after starting the job, and the status page cites reports of search and reporting being unavailable.

   Remediation

   • removed the reindex Job from the scheduler
   • let the nodes auto-heal back to healthy

   Lessons

   restartPolicy: Never on the pod template does not stop a Kubernetes Job from retrying: a Job replaces failed pods and may run more than once, so a job that can exhaust a node will keep moving to new nodes.

   View case

By root cause3