Cases2 cases

1 to 2 of 2

1. 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
2. IL-0241 A database shard could not elect a new primary, and the Stripe API failed most requests in two periods on one day

   Stripe2019

   Symptoms

   On 10 July 2019 the Stripe API failed a substantial majority of requests twice, from 16:35 to 17:02 UTC and from 21:14 to 22:47 UTC. In the first period automated monitoring detected within a minute that a database cluster had failed to elect a new primary.

   High error rateFailover failureWrites rejectedIntermittent timeoutsSustained high node CPU

   Trigger

   The primary node of a database cluster failed at 16:35 UTC while one shard had several stalled nodes. The second period was set off by rolling the election protocol back to the previous minor version at 20:42 UTC.

   Root cause

   Software defect

   Component

   DatabaseDatabase failover manager

   Contributing factor

   Change with a delayed effectMonitoring blind spotExcessive blast radiusHarmful manual actionUnverified assumptionOverlapping failures

   Detection

   Automated monitoring detected the failed election within a minute, and the team was alerted at 16:36 UTC. The second period was announced at 21:14 UTC by automated alerts that some shards in the cluster were unavailable.

   Remediation

   • Restarted all nodes of the database cluster, which forced a successful primary election
   • Updated the production configuration and restarted the database cluster's nodes
   • Ramped traffic back up, putting services for user-initiated API requests first

   Action items

   • Alert when nodes stop reporting their replication lag
   • Alert when a shard enters a state that could trigger the election fault
   • Work with the database maintainers on a fix for the underlying fault
   • Add circuit breaking on failed operations to particular database clusters
   • Use fault isolation to contain the impact of a single failed shard
   • Limit the resources used by clients that keep retrying failed requests
   • Add procedures and tooling that make rapid configuration changes during incident response safer

   View case

By root cause1