Cases4 cases

1 to 4 of 4

1. IL-0217 Honeycomb's main MySQL database locked up under cache-miss reads and ran out of connections, a 68-minute total outage

   Honeycomb2023

   Symptoms

   Hours after a routine switch between two clusters of Retriever, Honeycomb's storage and query engine, late on 24 July 2023, the internal SLO for Shepherd, the ingest service, began to burn slowly. The performance was only marginally slow and inconsistent, with no good explanation, so engineers treated it as minor and left it for the morning.

   High latencyDatabase overloadLock contention or deadlockService outage

   Trigger

   Late on 24 July Honeycomb switched between two Retriever clusters to avoid a subtle bug, and the new cluster stopped writing the last-written timestamps the schema cache depends on. On the morning of 25 July a feature flag flip meant to move those writes back stopped the remaining writes entirely.

   Root cause

   Software defect

   Component

   DatabaseIn-memory cache

   Contributing factor

   Dependence on a warm cacheIgnored early signalHarmful manual actionUnverified assumption

   Detection

   Hours after the cluster switch, the internal SLO for the ingest service Shepherd started burning slowly. Engineers left it for the morning, when a chart showed that Retriever's database calls updating the last-written records had stopped.

   Remediation

   • Flipped only the writes back to the other Retriever cluster with a feature flag, which did not fix the problem
   • Set up circuit breakers that denied all Shepherd traffic with a 5xx error to protect the database
   • Failed the hard-locked database over to a replica
   • Restarted all Retriever hosts so they could provide data for the cache
   • Set the last-written timestamp of every schema with traffic in the last day to now, so the schema store reloaded the cache
   • Removed the circuit breaker and restarted the query hosts that failed internal checks

   Action items

   • Finish the Retriever cluster migration and remove all code that could disable writes behind the cache
   • Explore ways to strengthen the cache further
   • Lower the contention put on the database during update storms from new schemas
   • Stabilise the performance costs of some operations in the system
   • Study instrumentation and experimentation that detect and handle such edge cases better

   Lessons

   Locally good decisions combined into the outage: the query engine wrote the timestamps because it already saw all the data, the shared cache helped scale ingest and untangle the main database, the cluster switch was done to avoid a bug, frequent deploys hid the feature flag bug, pausing deploys during the investigation kept hosts from restarting, and past near-misses made the team expect a slow failover. Feature flags, frequent deploys, pausing deploys during incidents and learning from near-misses all contributed, although they are among the practices Honeycomb trusts most. For a rare lockup of this kind a database failover is the fastest recovery. Several recent Honeycomb incidents started from trying to avoid a subtle bug.

   View case
2. IL-0213 Roblox was down for 73 hours because key-value writes on its single Consul cluster slowed from under 300 ms to 2 seconds

   Roblox2021

   Symptoms

   On the afternoon of 28 October 2021 (13:37 PST), Vault performance at Roblox degraded and a single Consul server showed high CPU load. Consul cluster metrics showed that writes to its key-value (KV) store had a 50th percentile latency of 2 seconds instead of the usual under 300 ms. Players were not yet affected.

   High latencySustained high node CPUService outageMonitoring outage

   Trigger

   On 27 October at 14:00, one day before the outage, Roblox enabled the Consul streaming feature on the backend service responsible for traffic routing and, to prepare for end-of-year traffic, added 50% more nodes to traffic routing.

   Root cause

   Software defect

   Component

   Coordination and key-value storeService discoveryIn-memory cache

   Contributing factor

   Single point of failureOverlapping failuresMonitoring blind spotHarmful manual actionChange with a delayed effect

   Detection

   Roblox engineers began investigating when Vault performance degraded and a single Consul server had high CPU load, before players were affected. Consul cluster metrics then showed elevated KV write latency.

   Remediation

   • Replaced one Consul cluster node on the suspicion of degraded hardware
   • Replaced all Consul nodes with 128-core machines with faster NVMe SSDs
   • Shut down the Consul cluster and reset its state from a snapshot taken at the start of the outage, with access blocked by iptables
   • Blocked external traffic, disabled all non-essential Consul use, scaled services down and lowered health check frequency from 60 seconds to 10 minutes
   • Moved the Consul cluster back to 64-core servers
   • Disabled the streaming feature for all Consul systems, which brought KV write latency back to 300ms
   • Kept the slow Consul servers from staying elected as leader
   • Redeployed the caching system and fixed its wrong scheduling data and failed deployments
   • Restarted the stopped services at the right capacity levels
   • Let players back in with DNS steering in roughly 10% steps while checking database load, cache performance and stability

   Action items

   • HashiCorp builds laboratory benchmarks and scale tests that reproduce the streaming contention
   • HashiCorp improves the design of streaming so it does not contend under extreme load
   • Compact the BoltDB database with existing BoltDB tooling
   • Remove the circular dependency between the telemetry systems and Consul
   • Extend telemetry with targeted alerts on Consul and BoltDB performance and with visibility into traffic between services and Consul
   • Run the backend in an additional, geographically distinct data center and in multiple availability zones
   • Split critical services into their own dedicated Consul clusters
   • Move data that core services keep directly in the Consul KV store to a more suitable storage system
   • Delete obsolete KV data from Consul
   • Upgrade to a Consul version that replaces BoltDB with bbolt, in Q1
   • Redesign cache deployment so the cache system can be brought up quickly from a standing start
   • Deploy Nomad enhancements that make it easier to turn up large jobs after a long period of unavailability
   • Use mechanisms for faster machine configuration changes
   • Reintroduce streaming only after a new implementation has been tested at Roblox's scale and workload

   Lessons

   Two mostly unrelated problems deep inside Consul, contention in streaming under very high read and write load and BoltDB freelist growth, were hard to diagnose, and that difficulty was the main reason the outage lasted so long. A single Consul cluster serving every workload made the impact larger, and monitoring that depended on Consul hid the cause when it was needed most. Faster hardware did not help and may have hurt, because the contention was worse on NUMA machines. HashiCorp had benchmarked streaming at a similar scale, but the problem needed both a large number of streams and a high churn rate. Restarting a platform from a fully stopped state, including warming caches, took a long time and needed tools built for bootstrapping rather than incremental change.

   View case
3. IL-0296 Transloadit's us-east Redis cluster lost queued jobs and heartbeats, raising WORKER_JOB_ERROR errors for 81 customers

   Transloadit2016

   Symptoms

   Reports came in of more WORKER_JOB_ERROR errors, which tell Transloadit there is a communication problem between its Uploaders, the machines that receive files and orchestrate Assemblies, and its Drones, the machines that run encoding tasks. Many Assemblies became much slower because retries with exponential back-off kicked in, and a few could not be replayed.

   High error rateHigh latencyLoad imbalance

   Trigger

   A steep increase in HLS traffic, together with more Drones online processing the same volume faster, which put more load on the central Redis nodes.

   Root cause

   Capacity exhaustion

   Component

   In-memory cacheDatabase replication

   Contributing factor

   Monitoring blind spotUnverified assumptionHarmful manual action

   Detection

   Reports of an increased number of WORKER_JOB_ERROR errors.

   Remediation

   • Re-added lost messages and relaxed heartbeat monitoring as a stopgap
   • Upgraded Redis to the latest stable version
   • Doubled the Redis cluster to six nodes and then added three more, which made it worse
   • Brought the number of read replicas down to one, which cut traffic and failure rates
   • Capped how far Drones scale up in each region

   Action items

   • Add logging and alarms for approaching saturation
   • Find ways to decentralise the traffic between the Redis nodes and the Drones and Uploaders
   • Add sharding per Uploader on top of the sharding per region

   Lessons

   To handle more traffic Transloadit had to use fewer Redis nodes, not more: each added replica multiplied the replication traffic from the single leader that takes every write, so scaling up the number of nodes was the wrong call. Bandwidth had not been among the monitored Redis metrics.

   View case
4. IL-0215 Twilio's billing Redis master lost all balances and went read-only, and auto-recharge charged customer cards repeatedly

   Twilio2013

   Symptoms

   At 2:39 AM PDT (9:39 UTC) on 18 July 2013, services that relied on the redis-master of Twilio's billing system began to fail because the load on its host had become extreme. At 2:42 AM PDT monitoring alerted the on-call engineers to a failure in the Redis cluster.

   Database overloadData lossWrites rejected

   Trigger

   At 1:35 AM PDT on 18 July a loss of network connectivity disconnected all billing redis-slaves from the redis-master at once, and all of them requested a full synchronization at the same time. On-call engineers then restarted the overloaded redis-master.

   Root cause

   Configuration change

   Component

   In-memory cacheDatabase replication

   Contributing factor

   Retry stormHarmful manual actionUnsafe failure behaviour

   Detection

   At 2:42 AM PDT monitoring alerted the on-call engineers to a failure in the Redis cluster. At 3:28 AM PDT the billing system monitoring reported the anomalous charges, and at 12:36 PM PDT monitoring detected the recurrence.

   Remediation

   • Restarted the redis-master to relieve the high load, which loaded a wrong configuration
   • Took the billing system offline to stop further credit card charges
   • Restored service to all suspended accounts
   • Restored the lost account balances from the independent double-bookkeeping record in the relational datastore
   • Took billing offline again after the recurrence and reactivated the newly suspended accounts
   • Refunded all erroneous credit card charges
   • Gave all affected accounts a credit of 10% of their last 30 days of spend
   • Replaced the original Redis cluster and corrected the redis-master configuration
   • Activated the billing system progressively for groups of customer accounts

   Action items

   • Disable Redis restarts on the redis-master and recover the master by pivoting a slave instead
   • Add fail-safes so the billing system neither suspends accounts nor charges cards when balances are missing or cannot be written
   • Validate the billing system against the double-bookkeeping databases in real time

   Lessons

   Losing the in-flight balances and the ability to write balances at the same time exposed a flaw in the auto-recharge system: it failed in a dangerous way, charging cards and suspending accounts instead of stopping.

   View case

By root cause3