Cases18 cases

13 to 17 of 18

1. IL-0206 Consul agents in GitLab.com's GKE pods were repeatedly OOM-killed, so the Rails pods lost database replica discovery and sent every query to the primary

   GitLab2022

   Symptoms

   On 2022-04-25 the Sidekiq Apdex (the share of background jobs that finish within a target time) dropped, and an engineer declared an incident in Slack at 15:13 UTC. Users saw intermittent slowness across GitLab.com, and asynchronous work such as pipeline creation and import and export took unusually long.

   High latencyDatabase overloadOOMKilledHigh error rate

   Trigger

   The Consul agent in GitLab.com's GKE pods had memory demand spikes that did not fit in the pods' 500MB memory limit, and the agent was killed again and again.

   Root cause

   Capacity exhaustion

   Component

   Container memory limitService discoveryApplication database

   Contributing factor

   –

   Detection

   An engineer declared the incident in Slack at 15:13 UTC after the Sidekiq Apdex dropped; the second event was found when monitoring detected an elevated error rate at 01:21 UTC on 2022-04-26.

   Remediation

   • planned at 16:13 to raise the per-user connection limit for the gitlab database user and the pgbouncer backend pools for the main database; changes were applied at 17:26 and confirmed at 17:47
   • raised the pod memory limits from 500MB to 1GB so the Consul agents survive memory spikes

   Action items

   • Monitor the Consul agents on GKE pods to see whether the 1GB memory limit keeps them healthy
   • Investigate the Consul memory demand spikes and what can be done beyond raising the hard memory limit

   Lessons

   When the local Consul agent that Rails uses for database service discovery keeps dying, Rails does not fail; it sends every query to the last known primary. What engineers saw was database saturation and slow background jobs, which pointed the investigation first at a recent pgbouncer pool change. The memory limit of the Consul agent in every pod therefore decided the load on the primary database.

   View case
2. 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
3. IL-0196 Adding a second Redis replica and lowering memory limits in one RedisFailover change restarted the only Redis pod that held data, and Sentinel promoted the empty new replica, erasing the data

   Palark2020

   Symptoms

   Right after an updated RedisFailover resource was rolled out, the large Redis database lost all its data and the service went down; the data had to be restored from a backup taken just before the change.

   Data lossService outage

   Trigger

   A RedisFailover change that raised the number of Redis replicas from one to two and, in the same rollout, lowered the containers' memory requests and limits from 22 GB to 19 GB.

   Root cause

   Software defect

   Component

   Application databasePersistent volumes and stateful setsContainer memory limit

   Contributing factor

   No staged rolloutSingle point of failureUnverified assumption

   Remediation

   • restored the Redis data from the backup prepared before the rollout

   Action items

   • Report the readiness probe flaw to the Redis Operator project and test the developers' fix
   • Reconsider the use of the Spotahome Redis Operator
   • Clean up the Redis databases and fix the application code that created the excess keys
   • Consider Redis built-in clustering and sharding

   Lessons

   Changing the replica count and the pod resources in one rollout turned a seemingly harmless change into a rolling restart of the only pod with data while the new replica was still empty. A readiness probe that only checks that Redis answers a ping does not tell whether a replica has finished syncing. Operators for critical stateful components need to be evaluated for maturity before production use, and the dump.rdb file should be backed up before any change to the Redis cluster.

   View case
4. 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
5. IL-0037 Image registry rate-limited a whole region after a crash-looping DaemonSet with imagePullPolicy Always pulled through three NAT instances

   Datadog2019

   Symptoms

   Users reported that their jobs were not starting, with some sign that image pulls were involved. Many pods across the cluster were in error states related to image pulls, while applications that were already running were not affected.

   Image pull failuresService outage

   Trigger

   A change to the permissions on a storage bucket that one application needed to read. The application could no longer start, and its DaemonSet, running on roughly a thousand nodes of the cluster, went into a crash loop with imagePullPolicy: Always, so every restart pulled the image again although it was already on the node.

   Root cause

   Capacity exhaustion

   Component

   Container registry and its credentialsNAT gatewayApplication databaseAdmission webhook

   Contributing factor

   Automation amplificationSingle point of failureUnknown limitHarmful manual actionExcessive blast radius

   Detection

   Users reported that their jobs were not starting; monitoring had not caught it first.

   Remediation

   • replaced the affected NAT instances so registry traffic came from new source IPs
   • reconfigured the static IPs and firewall rules so applications could reach the Cloud SQL database again

   Action items

   • Add an admission webhook that rejects floating image tags, starting with latest
   • Apply the image tag webhook to workload objects only, not to pods, after it blocked pods created by in-cluster controllers

   Lessons

   The speakers' takeaway is to be very careful with DaemonSets in large clusters: they say they had broken everything DaemonSets depend on, including DNS, Vault and the image repository, and that it is very easy to break things when managing them by the thousand. imagePullPolicy: Always made every crash-loop restart pull the image again. Sending all registry traffic through three NAT instances put the whole region behind three source IPs, which the registry rate-limited, and changing those IPs cut the connections to a critical database reached over the internet through them. An admission webhook on pods rejects the pods that controllers in the cluster create, where no user sees the error.

   View case

By root cause8