The worker catalog
This is the complete list of configurable background workers a cell can run, grouped by the service that owns them. There are 302 worker declarations across nine services (a handful of names, the metering flushers, the metrics publisher, the lifecycle listener, the crypto poller, recur in most services; those are factored out into the fleet-common workers table and are not repeated per service below).
Every worker is configured under
orchestrator.services.<service>.workers.<name> in the service configuration.
The Worker column is the literal config key an operator sets enabled and
interval_sec on. The Capability column is the governed capability
the worker serves. Enabled by default and Interval are the shipped
values; a bold No marks a worker that ships off and must be turned on
deliberately (see Declared-but-inert and off-by-default workers).
Worker keys and capability ids are the exact configuration identifiers, shown verbatim so this page is usable as an operator reference. Descriptions use public product names throughout; a few keys retain engineering vocabulary ("substrate" is retained in code by standard, surfaced to customers as Alignment).
System
The platform's spine: fabric membership and liveness, replication engine and topology, licensing, governance (Alignment) enforcement, session lifecycle, and Fleet Releases roll-out control.
Every cell in this service also runs the fleet-common workers. Listed below: 56 service-specific workers.
| Worker | Capability | Enabled by default | Interval | What it does |
|---|---|---|---|---|
substrate_cancer_scan | core.governance | Yes | 15m | Alignment (substrate) scan for runaway coupling growth ('cancer'). |
substrate_coupling_drift | core.governance | Yes | 30m | Alignment scan for coupling drift between components. |
automation_agent_dispatch | agents.workflow_dispatch | Yes | 5s | Dispatches Automations-driven agent workflows. |
agent_trigger_tick | agents.trigger | Yes | 30s | Fires Orchestration agent triggers. |
audit_retention | data.soft_delete | Yes | 6h | Purges soft-deleted directory rows past retention (the soft-delete purge worker). |
entity_retention | data.soft_delete | Yes | 6h | Purges soft-deleted entities past retention. |
archive_retention | data.soft_delete | Yes | 6h | Sweeps archived workspaces past retention. |
license_status_sweep | core.license | Yes | 60m | Sweeps license status. |
peer_bundle_forward | observability.metering | Yes | 5m | Forwards metering bundles to peers. |
cap_counter_rehydration | observability.metering | Yes | 15m | Rehydrates capability counters from the durable store. |
metering_retention | observability.metering | Yes | 60m | Enforces metering retention. |
node_log_retention | observability.retention | Yes | 6h | Enforces node-log retention. |
cohort_move_worker | core.migration | Yes | 30s | Moves cells between named cohorts. |
crypto_audit_witness | audit.tamper_evident | Yes | 60m | Witnesses the crypto audit chain. |
audit_chain_attestation | audit.tamper_evident | Yes | 60m | Attests the audit chain. |
audit_chain_attestation_selfcheck | audit.tamper_evident | Yes | 2h | Self-checks audit-chain attestation. |
crypto_audit_witness_selfcheck | audit.tamper_evident | Yes | 2h | Self-checks the crypto-audit witness. |
setup_runner | core.setup | Yes | continuous | Long-running first-boot setup driver (runs to completion, not periodic). |
fabric_metrics_publisher | metrics.federation | Yes | 15s | Publishes fabric-level metrics. |
replication_metrics_publisher | metrics.federation | Yes | 15s | Publishes replication metrics. |
cell_heartbeat | core.fabric | Yes | continuous | Long-running cell heartbeat to the fabric. |
service_group_reconcile | replication.service_group | Yes | 10s | Reconciles service-group membership continuously. |
membership_reconcile | replication.service_group | Yes | 30s | Reconciles cell membership continuously. |
cell_liveness | core.fabric | Yes | continuous | Long-running cell-liveness tracker. |
decommission_listener | core.lifecycle | Yes | continuous | Listens for and drives cell decommission. |
crypto_rotation | security.crypto_rotation | Yes | 24h | Daily crypto rotation driver. |
kek_forward_secret_ratchet | security.crypto_rotation | No | 90d | Forward-secret KEK ratchet. Off by default (90-day cadence when enabled). |
substrate_mission_expiry | core.governance | Yes | 60s | Expires Alignment mission grants. |
substrate_backfill | core.governance | Yes | 60m | Backfills Alignment (substrate) records. |
node_mind_consolidation | core.fabric | Yes | 5m | Consolidates the node-level fabric view ('node mind'). |
gateway_health_probe | core.fabric | Yes | 15s | Probes gateway health. |
capacity_reconcile | core.fabric | Yes | 60s | Reconciles fabric capacity accounting. |
license_db_watcher | core.license | Yes | 30s | Watches the license database for changes. |
license_file_watcher | core.license | Yes | 30s | Watches the license file for changes. |
link_quality_recorder | core.fabric | Yes | 30s | Records inter-cell link quality (Radio). |
gateway_sync | core.fabric | Yes | 60s | Syncs gateway configuration. |
session_revocations | core.session_revocation | No | continuous | Session-revocation producer. Off by default (event-driven revocation path). |
session_revocation_consumer | core.session_revocation | No | continuous | Session-revocation consumer. Off by default. |
session_cleanup | core.session_revocation | Yes | 5m | Cleans up expired sessions (the enabled session-hygiene worker). |
endpoint_reaper | core.fabric | Yes | 60s | Reaps stale service endpoints. |
lease_sweeper | core.fabric | Yes | 30s | Sweeps expired fabric leases. |
volume_health | core.fabric | Yes | 60s | Checks volume health. |
replication_engine | replication.engine | Yes | 2s | The replication engine tick. |
replication_topology | replication.engine | Yes | 60s | Maintains replication topology. |
replication_health | replication.health | Yes | 30s | Checks replication health. |
anti_entropy | replication.anti_entropy | Yes | 5m | System-store anti-entropy. |
replication_repair | replication.anti_entropy | Yes | 60m | Repairs replication gaps. |
connection_discovery | core.fabric | Yes | 60s | Discovers peer connections. |
db_size_meter | observability.metering | Yes | 5m | Meters per-database size. |
fleet_rollout | observability.fleet | Yes | 15s | Drives Fleet Releases roll-out across cohorts. |
fleet_drift | observability.fleet | Yes | 30s | Detects fleet version/config drift. |
fleet_eligibility | observability.fleet | Yes | 60s | Maintains the fleet roll-out eligibility cache. |
fleet_auto_remediation | observability.fleet | Yes | 60s | Auto-remediates unhealthy fleet members. |
fleet_health | observability.fleet | Yes | 60s | Aggregates fleet health. |
fleet_version_distribution | observability.fleet | Yes | 60s | Tracks fleet version distribution. |
fleet_alert_evaluator | observability.fleet | Yes | 60s | Evaluates fleet alerts. |
Data
The Data Platform's background plane: lakehouse and catalog upkeep, CRDT convergence, vector/search/document maintenance, clean rooms, Automations dispatch, Knowledge promotion, and consulting workflows.
Every cell in this service also runs the fleet-common workers. Listed below: 59 service-specific workers.
| Worker | Capability | Enabled by default | Interval | What it does |
|---|---|---|---|---|
lakehouse_anti_entropy | data.lakehouse | Yes | 5m | Repairs divergence in lakehouse table state across cells. |
liquid_cluster_sweep | data.lakehouse | Yes | 10m | Sweeps and re-clusters lakehouse tables for query locality. |
stream_wal_checkpoint | data.stream | Yes | 5m | Checkpoints the streaming write-ahead log so it does not grow without bound. |
clone_ttl | data.clone | Yes | 60s | Expires clones created with a TTL so the clone estate does not grow without bound. |
crdt_anti_entropy | replication.anti_entropy | Yes | 30s | Drives CRDT anti-entropy so replicas converge. |
catalog_leaf_anti_entropy | catalog.leaf_replication | Yes | 60s | Reconciles the signed catalog leaf-inventory projection across cells (Discovery & Mapping). |
crdt_tombstone_gc | data.crdt_compaction | Yes | 2m | Garbage-collects settled CRDT tombstones after the safety window. |
ttl_scanner | data.ttl | Yes | 0.1s | High-frequency scan that expires TTL'd rows within a small time budget per tick. |
snapshot | data.snapshot | Yes | 5m | Takes periodic data snapshots. |
asset_prognosis | data.asset_prognosis | Yes | 60m | Runs Predictive Maintenance asset-prognostics signal processing over device telemetry. |
credential_lifecycle | data.governance | Yes | 60m | Rotates and expires catalog data-source credentials. |
catalog_search_reindex | data.governance | Yes | 5m | Reindexes the catalog search index. |
catalog_fts_reindex | data.governance | Yes | 5m | Rebuilds the catalog full-text search index. |
wal_compaction | data.compaction | Yes | 60s | Compacts the data write-ahead log. |
eviction_scanner | data.eviction | Yes | 0.1s | High-frequency cache/eviction scan. |
replication_sync | replication.crypto | No | 5s | Legacy replication sync stub. Off by default and inert: kept disabled so re-enabling without an implementation fails the build. See the inert section. |
graph_replication | replication.graph | Yes | 5s | Replicates graph-store changes across cells. |
structures_lifecycle_listener | core.lifecycle | Yes | 5s | Reacts to lifecycle events for data structures. |
transaction_recovery | data.txn_recovery | Yes | 30s | Recovers in-doubt transactions after a crash. |
catalog_health | data.catalog | Yes | 2m | Checks catalog store health. |
consulting_retention | data.retention | Yes | 60m | Enforces retention on consulting engagement data. |
siem_lake_retention | data.retention | Yes | 60m | Enforces retention on the SIEM data lake. |
consulting_assessment_runs | data.consulting | Yes | 5s | Drains and advances queued consulting assessment runs. |
consulting_recurring_schedules | data.consulting | Yes | 5m | Materializes due consulting recurring schedules. |
consulting_accepted_risk_expiry | data.consulting | Yes | 60m | Expires accepted-risk findings past their acceptance window. |
catalog_sync | data.catalog | Yes | 5m | Syncs catalog metadata. |
lineage_enrichment | data.lineage | Yes | 30s | Enriches data-lineage edges as new events arrive. |
warehouse_lineage_poll | data.lineage | Yes | 30m | Polls connected warehouses for lineage. |
vector_maintenance | data.vector | Yes | 5m | Maintains vector indexes (merges, compaction). |
vector_rebuild_worker | data.vector | Yes | 30m | Rebuilds vector indexes for scale-out shards. |
structures_maintenance | data.compaction | Yes | 5m | Compacts and maintains structured stores. |
materialized_view_refresh | data.materialized_views | Yes | 5m | Refreshes materialized views. |
protection_rebuild | replication.protection | Yes | 5m | Rebuilds erasure/replication protection for scale-out data. |
drift_retrain | data.ml_retrain | No | 60m | Retrains models on detected data drift. Off by default (opt-in ML). |
search_rebuild_worker | data.search | Yes | 30m | Rebuilds search indexes for scale-out shards. |
document_maintenance | data.document | Yes | 10m | Maintains the document store. |
search_maintenance | data.search | Yes | 10m | Maintains search indexes. |
feature_materialization | data.feature_store | Yes | 30m | Materializes feature-store features on schedule. |
streaming_eval | data.streaming | Yes | 5s | Evaluates streaming query state. |
capacity_stats_meter | observability.metering | Yes | 60s | Meters per-cell data capacity statistics. |
engine_stats_sync | data.engine_stats | Yes | 2m | Syncs query-engine statistics for the planner. |
meta_intel_index_sync | data.meta_intelligence | No | 2m | Metadata-intelligence index sync. Off by default and inert: no live provider. See the inert section. |
meta_intel_quality_compute | data.meta_intelligence | No | 5m | Metadata-intelligence quality scoring. Off by default and inert. See the inert section. |
meta_intel_enrichment | data.meta_intelligence | No | 10m | Metadata-intelligence enrichment scan. Off by default and inert. See the inert section. |
meta_intel_usage_rollup | data.meta_intelligence | No | 60m | Metadata-intelligence usage rollup. Off by default and inert. See the inert section. |
meta_intel_meter_flush | observability.metering | Yes | 60s | Flushes metadata-intelligence metering (a real flush, unlike the four inert meta-intelligence workers). |
cleanroom_kek_rotation | data.cleanroom | Yes | 60m | Rotates clean-room key-encryption keys. |
cleanroom_expiration | data.cleanroom | Yes | 5m | Expires clean rooms past their lifetime. |
cleanroom_anomaly_sweep | data.cleanroom | Yes | 5m | Sweeps clean rooms for access anomalies. |
cleanroom_substrate_classify | data.cleanroom | Yes | 15m | Runs Alignment (substrate) classification over clean-room content. |
automation_tick | data.streaming | Yes | 30s | Evaluates reactive Automations rules and enqueues matches. |
automation_forge_dispatch | data.streaming | Yes | 5s | Dispatches queued Automations-driven Data Modeling jobs. |
workspace_anti_entropy | replication.failover | Yes | 5m | Reconciles workspace state across cells for failover readiness. |
failover_detector | replication.failover | Yes | 5s | Detects a failed home cell and triggers standby election. |
rebalance_membership | replication.rebalance | Yes | 30s | Rebalances data ownership on membership change. |
knowledge_coherence | cognition.knowledge | No | 5m | Knowledge (graph) coherence pass. Off by default. See the inert section note on cognition.knowledge. |
knowledge_promotion | cognition.knowledge | Yes | 60s | Promotes candidate Knowledge entities to canonical under lease. |
knowledge_coherence_stitch | cognition.knowledge | Yes | 5m | Stitches the unified Knowledge graph across cohorts. |
knowledge_tier | cognition.knowledge | No | 5m | Tiers Knowledge entities hot/warm/cold by age. Off by default (requires an explicit cell id). |
Security
Security & Operations background work: crypto health and rotation, threat detection and intelligence, posture and compliance scoring, event ingest and SIEM forwarding, and the offense-capable workers that ship off by default.
Every cell in this service also runs the fleet-common workers. Listed below: 43 service-specific workers.
| Worker | Capability | Enabled by default | Interval | What it does |
|---|---|---|---|---|
peer_audit_reconcile | security.audit_reconcile | Yes | 2m | Reconciles audit chains with peer cells. |
zeroize_trigger_monitor | security.tactical_zeroize | Yes | 60s | Scaffold, not an enforced control. On every profile shipping today it is a safe no-op: most profiles declare no zeroize triggers and no tactical hardware provider is registered, so nothing can arm and nothing fires. Where a profile declares a trigger that no hardware backs, it logs an error rather than acting. Auto-zeroize on disconnect is declared/scaffolded, not a working guarantee. |
expiry_monitor | security.crypto_health | Yes | 60s | Warns on impending certificate/key expiry. |
gossip_sync | security.crypto_rotation | Yes | 30s | Gossips crypto-rotation state with peers. |
vault_integrity | security.crypto_health | Yes | 5m | Checks secret-vault integrity. |
rotation_coordinator | security.crypto_rotation | Yes | 60m | Coordinates scheduled crypto rotation. |
federation_health | security.federation | Yes | 60m | Checks security-federation health. |
dlp_scanner | security.posture | No | 6h | Scans object stores for data-loss-prevention hits. Off by default (opt-in, resource-heavy). |
security_event_ingest | security.event_ingest | Yes | 30s | Ingests security events into the security graph. |
siem_forward | security.event_ingest | Yes | 30s | Forwards security events to an external SIEM. |
ct_submit | security.crypto_health | No | 5m | Submits certificates to Certificate Transparency logs. Off by default (needs egress). |
acme_public_cert_renewal | security.crypto_rotation | Yes | 12h | Renews public ACME certificates. |
ocsp_stapling_prefetch | security.crypto_health | Yes | 60m | Prefetches OCSP staples for served certificates; the cached response is stapled onto the cell's TLS listeners and served in the handshake (fail-safe to unstapled when a leaf has no current staple). |
crypto_drift_scan | security.crypto_health | Yes | 15m | Scans for cryptographic configuration drift. |
security_graph_builder | security.graph | Yes | 2m | Builds the security relationship graph. |
detection_engine | security.detection | Yes | 60s | Runs detection rules over ingested events. |
baseline_builder | security.baseline | Yes | 6h | Builds behavioral baselines for detection. |
posture_scorer | security.posture | Yes | 15m | Scores security posture. |
sla_monitor | security.posture | Yes | 5m | Monitors security SLAs. |
ct_auditor | security.crypto_health | Yes | 60m | Audits Certificate Transparency observations. |
compliance_monitor | security.posture | Yes | 60m | Evaluates compliance controls. |
identity_graph_sync | security.federation | Yes | 10m | Syncs the identity graph across federation. |
security_agent_runner | security.agent | Yes | 60s | Runs scheduled security agents. |
intel_sharing | security.threat_intel | Yes | 60s | Shares threat intelligence with peers. |
threat_aggregation | security.threat_intel | Yes | 5m | Aggregates threat signals. |
response_executor | security.response | Yes | 10s | Executes automated response actions. |
swarm_aggregator | security.swarm | Yes | 30s | Aggregates swarm-defense signals across the fleet. |
device_attestation | security.device_attestation | Yes | 15m | Intended to attest IoT device posture. Enabled but its tick is currently inert (see the inert section); the underlying posture surface is served live elsewhere. |
botnet_detection | security.detection | Yes | 30s | Detects botnet/C2 patterns (Threat Detection). |
indicator_propagation | security.threat_intel | Yes | 5s | Propagates threat indicators quickly across cells. |
consensus_engine | security.consensus | Yes | 10s | Runs the reputation-consensus protocol. |
mtd_rotation | cyberops.offensive | No | 10s | Moving-target-defense rotation. Off by default (offense-capable; enable only where authorized). |
active_defense_wall | cyberops.offensive | Yes | 5s | Adaptive active-defense wall (Data Security). Enabled. |
counter_botnet | cyberops.offensive | Yes | 60s | Counter-botnet operations (Counter-Threat Operations). Enabled. |
cyber_fires | cyberops.offensive | Yes | 10s | Cyber 'fires' authorization loop (Counter-Threat Operations). |
ct_intelligence | security.threat_intel | No | 60s | Certificate Transparency intelligence mining. Off by default. |
cognitive_defense | cyberops.cognitive | Yes | 30s | Cognitive-defense analysis loop. |
supply_chain | security.supply_chain | Yes | 60m | Scans software supply-chain posture. |
self_healing | security.self_healing | Yes | 15s | Applies self-healing remediations. |
rule_evolution | security.self_healing | Yes | 24h | Evolves detection rules on a daily cadence. |
autonomous_red_team | cyberops.offensive | No | 7d | Autonomous red-team exercises. Off by default (offense-capable; weekly when enabled). |
quantum_readiness | security.crypto_health | Yes | 24h | Assesses post-quantum crypto readiness. |
advanced_metrics | metrics.federation | Yes | 30s | Publishes advanced security metrics. |
Storage
Object and block durability: erasure and bitrot integrity, tiering, deduplication, replication and rebalance, quarantine repair, and lifecycle.
Every cell in this service also runs the fleet-common workers. Listed below: 26 service-specific workers.
| Worker | Capability | Enabled by default | Interval | What it does |
|---|---|---|---|---|
volume_compactor | storage.tiering_work | Yes | 60s | Compacts storage volumes. |
orphan_block_gc | storage.dedup | Yes | 2m | Garbage-collects deduplicated blocks no object references. |
multipart_cleanup | storage.dedup | Yes | 5m | Cleans up abandoned multipart uploads. |
storage_vacuum | storage.tiering_work | Yes | 5m | Vacuums storage metadata. |
replication | replication.crypto | Yes | 2s | Replicates objects to peer cells. |
peering_probe | storage.topology | Yes | 15s | Probes storage-peer topology and reachability. |
erasure_health | storage.integrity | Yes | 30s | Checks erasure-coded shard health. |
durability_audit | storage.integrity | Yes | 5m | Audits object durability against target. |
replica_lease_sweep | storage.tiering_work | Yes | 60s | Sweeps expired replica leases. |
bitrot_scan | storage.integrity | Yes | 5m | Rate-limited bitrot scan of stored data. |
anti_entropy | replication.anti_entropy | Yes | 60s | Repairs replica divergence. |
tiering | storage.tiering_work | Yes | 5m | Moves objects between storage tiers by policy. |
tier_transition_applier | storage.tiering_work | Yes | 5m | Applies queued tier transitions. |
lifecycle | core.lifecycle | Yes | 60m | Applies object lifecycle rules. |
change_stream | storage.change_events | Yes | 10s | Emits the storage change stream. |
integrity_scan | storage.integrity | Yes | 5m | Scans objects for integrity violations. |
quarantine_repair | storage.integrity | Yes | 10m | Repairs quarantined objects with bounded retries. |
cross_ns_processor | storage.change_events | No | 30s | Cross-namespace change processor. Off by default. |
key_lifecycle | storage.key_lifecycle | Yes | 5m | Rotates and expires storage encryption keys. |
event_dispatch | storage.change_events | Yes | 5s | Dispatches storage change events to subscribers. |
dedup_scan | storage.dedup | Yes | 60m | Scans for deduplication opportunities. |
service_group_migration | replication.rebalance | Yes | 60s | Migrates a service group's storage between cells. Now implemented (previously inert). |
protection_auto_scaler | storage.protection_autoscale | Yes | 60s | Auto-scales protection level under load. |
rebalance | replication.rebalance | Yes | 5m | Rebalances object placement across cells. |
clone_promotion | storage.clone | Yes | 30s | Promotes copy-on-write clones to independent objects. |
capacity_reconcile | replication.rebalance | Yes | 60s | Reconciles storage capacity accounting. |
Compute
The agent and job execution plane: kernel pools, warm pools, coding-agent and ML-training dispatch, git and CI triggers, and cloud price sync.
Every cell in this service also runs the fleet-common workers. Listed below: 21 service-specific workers.
| Worker | Capability | Enabled by default | Interval | What it does |
|---|---|---|---|---|
cognition_cost_flush | compute.kernel_reaper | Yes | 60s | Drains the shared Cognition cost collector into metering, using the canonical collector so no rate is lost. |
secret_rotation | compute.kernel_reaper | Yes | 12h | Monitors master-key age and warns when a key has aged past its window; observation only unless auto-rotate is explicitly requested. Self-gates when no key exists. |
kernel_reaper | compute.kernel_reaper | Yes | 30s | Reaps idle and expired compute kernels so the kernel pool does not leak capacity. |
warm_pool | compute.warmpool | Yes | 10s | Maintains the warm kernel pool so a job does not pay cold-start latency. |
library_sync | compute.library | Yes | 2m | Keeps the per-workspace code/library trees in sync for compute jobs. |
job_tick | compute.jobs | Yes | 5s | Drains the cell-local job queue and advances claimed jobs. |
ml_training_dispatch | compute.ml_training | Yes | 5s | Drains the ML-training queue: resolves a snapshot, fits, exports ONNX, node-signs, stores, and registers the model version. Default-enabled so training works out of the box. |
coding_run_dispatch | compute.jobs | Yes | 5s | Drives each claimed coding run through provision, environment bootstrap, self-repair, and settle. Fail-closed: an AI-off or git-disabled cell runs an idle no-op tick rather than a fabricated run. |
cloud_price_sync | observability.metering | Yes | 8.58358h | Seeds and syncs the per-cloud infrastructure price catalog so the cross-cloud cost comparator and placement router have live rates. Seed-and-curate only when no fetcher is wired. |
automation_dbt_dispatch | compute.jobs | Yes | 5s | Dispatches queued Data Modeling (dbt) transform-project runs. |
quota_sync | compute.quota | Yes | 30s | Reconciles compute quota accounting across the cell. |
agent_trigger | agents.trigger | Yes | 15s | Fires eligible agent triggers whose conditions have been met. |
git_sync | compute.git | Yes | 60s | Syncs bound workspace git repositories. |
ci_trigger | compute.ci | Yes | 30s | Evaluates and fires CI pipeline triggers. |
gguf_warmup | compute.warmpool | No | 10m | Pre-warms local GGUF model weights into memory. Off by default; a local-model-serving opt-in. |
gguf_idle_evictor | compute.warmpool | No | 60s | Evicts idle local GGUF models to reclaim memory. Off by default, pairs with the warmup worker. |
model_build_pipeline | compute.model_build | No | 5s | Runs the local Model Provisioning build pipeline (acquire, quantize/compile, sign, stage). Off by default. |
intelligence_maintenance_tick | agents.intelligence_maintenance | Yes | 60s | Periodic maintenance of the intelligence/agent subsystem. |
workspace_dataplane_reconcile | core.lifecycle | Yes | 5m | Reconciles workspace data-plane placement against desired state. |
entity_home_reconcile | core.lifecycle | Yes | 5m | Reconciles the home cell of entities after membership change. |
workspace_home_reassign | core.lifecycle | Yes | 5m | Reassigns a workspace's home cell when its current home is unfit. |
Metrics
The observability and metering engine: collection, rollups, retention, alerting, anomaly and FinOps scans, usage rollups, budgets, and chargeback.
Every cell in this service also runs the fleet-common workers. Listed below: 18 service-specific workers.
| Worker | Capability | Enabled by default | Interval | What it does |
|---|---|---|---|---|
otlp_push | metrics.federation | No | 60s | Pushes metrics to an external OTLP endpoint. Off by default (cloud/connected opt-in). |
collection_engine | observability.collection | Yes | 1s | High-frequency metrics collection loop. |
rollup_compute | observability.rollup | Yes | 60s | Computes metric rollups. |
retention_sweep | observability.retention | Yes | 60m | Enforces metrics retention. |
alert_evaluator | observability.alerting | Yes | 15s | Evaluates alert rules against live metrics. |
anomaly_scan | observability.anomaly | Yes | 5m | Scans metrics for anomalies. |
finops_anomaly_scan | observability.finops | Yes | 60m | Scans cost/usage for FinOps anomalies. |
audit_chain_rotation | audit.tamper_evident | Yes | 60m | Rotates the tamper-evident audit chain segment. |
health_synthesizer | observability.health | Yes | 30s | Synthesizes an overall health signal from component metrics. |
usage_rollup_5min | observability.metering | Yes | 5m | Rolls up metered usage on a 5-minute grain. |
usage_rollup_hourly | observability.metering | Yes | 60m | Rolls up metered usage hourly. |
usage_rollup_daily | observability.metering | Yes | 24h | Rolls up metered usage daily. |
usage_sync | observability.metering | Yes | 5m | Syncs usage records across cells. |
chargeback_recompute_drain | observability.metering | Yes | 5m | Drains and recomputes chargeback allocations. |
budget_evaluator | observability.metering | Yes | 5m | Evaluates budgets and raises breach signals. |
rate_card_sync | observability.metering | Yes | 60m | Syncs the billing rate card. |
knowledge_cost_rollup | observability.metering | Yes | 60m | Rolls up Knowledge subsystem cost. |
agent_metrics_rollup | observability.metering | Yes | 60m | Rolls up per-agent metrics. |
Messaging
The cross-cell delivery fabric (Messaging & Streaming): outbox retransmit, inbox apply, cursor and ack management, gossip membership, change replication, and stream maintenance.
Every cell in this service also runs the fleet-common workers. Listed below: 17 service-specific workers.
| Worker | Capability | Enabled by default | Interval | What it does |
|---|---|---|---|---|
asset_replication_drain | replication.crypto | Yes | 10s | Drains queued asset-blob replication to peer cells. |
retransmit_due | messaging.delivery | Yes | 0.25s | Retransmits due outbox messages under a cell-scoped lease, with bounded attempts. |
apply_inbox | messaging.xcell | Yes | 0.25s | Applies received cross-cell inbox messages. |
reassemble_fragments | messaging.fragments | Yes | 1s | Reassembles fragmented large messages. |
consumer_group_expiry | messaging.delivery | Yes | 10s | Expires stale consumer-group memberships. |
advance_cursors | messaging.delivery | Yes | 1s | Advances subscription cursors and auto-commits within lag thresholds. |
ack_timeout_watcher | messaging.delivery | Yes | 1s | Detects ack timeouts and requeues or dead-letters per policy. |
partition_migration | messaging.delivery | Yes | 5s | Migrates partitions between cells. |
route_refresh | messaging.routing | Yes | 2s | Refreshes cell-to-cell routing from fabric endpoint/policy tables. |
gossip_membership | messaging.xcell | Yes | 2s | Gossips membership and endpoint liveness with peer cells. |
replicate_changes | messaging.xcell | Yes | 1s | Streams replication-log changes for system, directory, and session stores. |
streaming_query_monitor | messaging.maintenance | Yes | 60s | Monitors streaming query health. |
pipeline_monitor | messaging.maintenance | Yes | 60s | Monitors streaming pipeline health. |
retention_sweeper | messaging.maintenance | Yes | 60s | Enforces subject and inbox/dead-letter/fragment retention. |
outbox_backoff_reset | messaging.delivery | Yes | 1s | Resets outbox backoff after a transport recovers. |
vacuum | messaging.maintenance | Yes | 5m | Vacuums/checkpoints the messages store on a maintenance cadence. |
capacity_reconcile | replication.rebalance | Yes | 60s | Reconciles messaging capacity and partition placement. |
Node
Node-level roll-up and reporting: fleet-facing governance attestation polls, license refresh, metering and exchange reports, edge site-artifact pull, and telemetry bridges.
Every cell in this service also runs the fleet-common workers. Listed below: 17 service-specific workers.
| Worker | Capability | Enabled by default | Interval | What it does |
|---|---|---|---|---|
gateway_metrics | metrics.federation | Yes | 10s | Publishes gateway metrics. |
npg_attestation_poll | core.governance | Yes | 5m | Polls for net-potential-gain governance attestations (Alignment). |
worker_policy_attestation_poll | core.governance | Yes | 5m | Polls for Worker Policy attestations so worker-governance decisions are verifiable. |
site_artifact_poll | edge.site_artifact | Yes | 15s | Pulls signed edge site artifacts for the node. |
security_bridge_sync | security.bridge_sync | Yes | 5s | Syncs the node's security bridge state. |
license_refresh | core.license | Yes | 60m | Refreshes the node license. |
metering_report | observability.metering | Yes | 60m | Reports metered usage upstream (connected deployments). |
exchange_settlement_report | observability.metering | Yes | 60m | Reports data-exchange settlement. |
exchange_access_request_pull | observability.metering | Yes | 60m | Pulls pending data-exchange access requests. |
exchange_offer_publish_report | observability.metering | Yes | 60m | Reports data-exchange offer publications. |
telemetry_incident_report | observability.telemetry | Yes | 5m | Forwards incident telemetry (connected deployments). |
telemetry_bug_report | observability.telemetry | Yes | 10m | Forwards bug-report telemetry. |
telemetry_a11y_report | observability.telemetry | Yes | 30m | Forwards accessibility telemetry. |
telemetry_perf_report | observability.telemetry | Yes | 10m | Forwards performance telemetry. |
telemetry_perf_detector | observability.telemetry | Yes | 60s | Detects performance regressions locally. |
telemetry_log_bridge | observability.telemetry | Yes | 30s | Bridges logs to the telemetry plane. |
telemetry_incident_bridge | observability.telemetry | Yes | 30s | Bridges incidents to the telemetry plane. |
Admin
The control-plane service: fabric-health probing plus the fleet-common publishers.
Every cell in this service also runs the fleet-common workers. Listed below: 1 service-specific workers.
| Worker | Capability | Enabled by default | Interval | What it does |
|---|---|---|---|---|
fabric_health_agent | observability.fabric_health | Yes | 15s | Probes every peer service's readiness endpoint, keeps an EWMA health signal, and writes fabric-health state (via API or DB). Setup-aware and self-excluding. |