Cases2 cases

1 to 2 of 2

1. IL-0290 GitLab.com's Redis Sidekiq cluster ran out of memory and lost quorum, causing 500 errors and a 12 million job backlog

   GitLab2026sev1

   Symptoms

   From 17:12 UTC on 12 May 2026, error rates rose at the same time across GitLab.com's web, Git, API and CI runner services. Users got frequent 500 errors, failed or stuck CI pipelines, authentication failures and unreliable Git access, and a Prometheus Alertmanager alert opened the incident at 17:15.

   High error rateMemory pressureThroughput degradation

   Trigger

   Earlier on 12 May, as a mitigation for a previous incident, INC-10084, a Sidekiq queueing apdex SLO violation on shard catchall-b, jobs of AuditEvents::AuditEventStreamingWorker were deferred. The deferred audit event jobs then piled up in the Redis Sidekiq cluster.

   Root cause

   Capacity exhaustion

   Component

   In-memory cache

   Contributing factor

   Change with a delayed effectPages not deliveredLimited debug access

   Detection

   A Prometheus Alertmanager alert reported the incident at 17:15:10 UTC, three minutes after impact began. A follow-up asks why the redis-sidekiq memory alert did not page, and the review says some signals did not reach the right on-call people.

   Remediation

   • Restarted and force-reset the two unresponsive Redis Sidekiq nodes
   • Stopped each Redis Sidekiq VM in turn and resized it to c2-standard-16, doubling memory and CPU
   • Re-enabled AuditEventStreamingWorker instead of deferring it

   Action items

   • Investigate why the redis-sidekiq memory alert did not page
   • Establish the root cause, failure mode and remediation status of the WebHookWorker poison pill
   • Give mitigations that disable safety limits an explicit owner and a deadline for re-evaluating them
   • Stop AuditEvents::AuditEventStreamingWorker from causing recurring incidents
   • Create a runbook for AuditEventStreamingWorker
   • Address PG connection pool saturation

   Lessons

   Broad degradation across web, Git, authentication, CI runners, Sidekiq and the internal API made it hard to settle on one cause and explain it clearly. The customer symptom of pipelines stuck in created was hard to measure in real time, and the dashboards that would have answered it were not obvious to on-call responders. Some alerts did not reach the right on-call people, and hosts that hung on login slowed diagnosis and made changes riskier. Redis-sidekiq under memory saturation forces choices between availability and data loss that are hard to make during an incident and call for guardrails and design changes.

   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

By root cause2