Concept guides
ReadGuideTrackTopicLevelTime
Rate limitingMediumCodingCapacity and trafficCodingCapacity and trafficMedium3 minThe storage stackMediumCodingCapacity and trafficCodingCapacity and trafficMedium4 minMissing rate limitingMediumCodingCapacity and trafficCodingCapacity and trafficMedium3 minModelling the resources before the endpointsHardCodingCapacity and trafficCodingCapacity and trafficHard7 minHot partition saturationHardCodingCapacity and trafficCodingCapacity and trafficHard3 minFalse sharing (cache contention)HardCodingCapacity and trafficCodingCapacity and trafficHard3 minGPU memory fragmentationHardCodingCapacity and trafficCodingCapacity and trafficHard3 minInference GPU autoscalerHardCodingCapacity and trafficCodingCapacity and trafficHard3 minInference GPU OOM (out of memory)HardCodingCapacity and trafficCodingCapacity and trafficHard3 minLatency from a failing disk (triage, contain, root-cause)HardCodingCapacity and trafficCodingCapacity and trafficHard6 minShadow traffic overloadHardCodingCapacity and trafficCodingCapacity and trafficHard3 minThrottling and 429s on object storageHardCodingCapacity and trafficCodingCapacity and trafficHard3 minWAL, tombstones, and read amplificationHardCodingCapacity and trafficCodingCapacity and trafficHard4 min
Configuration errorsEasyCodingDependencies and deploysCodingDependencies and deploysEasy3 minEnvironment misconfigurationEasyCodingDependencies and deploysCodingDependencies and deploysEasy3 minBad data deployMediumCodingDependencies and deploysCodingDependencies and deploysMedium3 minCanary deploy and the bad versionMediumCodingDependencies and deploysCodingDependencies and deploysMedium3 minCold storage retrieval timeoutsMediumCodingDependencies and deploysCodingDependencies and deploysMedium3 minDependency CVEsMediumCodingDependencies and deploysCodingDependencies and deploysMedium3 minFeature store timeoutsMediumCodingDependencies and deploysCodingDependencies and deploysMedium3 minModel serving cold startMediumCodingDependencies and deploysCodingDependencies and deploysMedium3 minModel serving dependency fallbackMediumCodingDependencies and deploysCodingDependencies and deploysMedium3 minModel serving registry outageMediumCodingDependencies and deploysCodingDependencies and deploysMedium4 minTLS certificate expiryMediumCodingDependencies and deploysCodingDependencies and deploysMedium5 minTokenizer version mismatchMediumCodingDependencies and deploysCodingDependencies and deploysMedium3 minDeploys, rollouts & flagsMediumCodingDependencies and deploysCodingDependencies and deploysMedium3 minDependency & vendor failuresMediumCodingDependencies and deploysCodingDependencies and deploysMedium3 minML canary: the NaN trapHardCodingDependencies and deploysCodingDependencies and deploysHard3 minTrain-serve skewHardCodingDependencies and deploysCodingDependencies and deploysHard3 min
Database disk saturation (IOPS)MediumCodingDatabases in productionCodingDatabases in productionMedium3 minDatabase unreachableMediumCodingDatabases in productionCodingDatabases in productionMedium3 minLLM output contract breaksMediumCodingDatabases in productionCodingDatabases in productionMedium3 minDatabase incidentsMediumCodingDatabases in productionCodingDatabases in productionMedium3 minDatabase connection pool exhaustionMediumCodingDatabases in productionCodingDatabases in productionMedium3 minDatabase index bloatHardCodingDatabases in productionCodingDatabases in productionHard3 minLayer 4 (TCP) load balancingHardCodingDatabases in productionCodingDatabases in productionHard3 minQuery of deathHardCodingDatabases in productionCodingDatabases in productionHard3 minDatabase autovacuum falling behindHardCodingDatabases in productionCodingDatabases in productionHard3 minDatabase index corruptionHardCodingDatabases in productionCodingDatabases in productionHard3 minDatabase runaway migrationHardCodingDatabases in productionCodingDatabases in productionHard3 minFeature pipeline data qualityHardCodingDatabases in productionCodingDatabases in productionHard3 minHot partitions and shardingHardCodingDatabases in productionCodingDatabases in productionHard3 minPartition maintenance gapsHardCodingDatabases in productionCodingDatabases in productionHard3 minTCP: the reliable byte streamHardCodingDatabases in productionCodingDatabases in productionHard4 minVector index stalenessHardCodingDatabases in productionCodingDatabases in productionHard3 minWhen a node goes down, its shards re-homeHardCodingDatabases in productionCodingDatabases in productionHard4 min
A restore that stallsMediumCodingIncident responseCodingIncident responseMedium4 minCDN time-to-live (TTL)MediumCodingIncident responseCodingIncident responseMedium3 minCold storage SLAsMediumCodingIncident responseCodingIncident responseMedium3 minContent delivery networksMediumCodingIncident responseCodingIncident responseMedium3 min