Cluster Settings
Cluster settings control correlator behavior and are the same for all nodes. A configured value is replicated across the entire cluster.
Overriding Settings
By default, all parameters use their built-in values.
To specify a custom parameter value:
- Click
Override Setting - In
Select Setting, select the required parameter. Its current default value is displayed next to its name - Enter the parameter value
- Click
Addto apply the value, orCancelto close the form without changes
The parameter appears in a separate table of overridden settings. Resetting a parameter in this table restores its built-in value.

When a Value Takes Effect
The Restart column in the tables below shows how a parameter takes effect:
no- the value is applied to the running cluster as soon as it is savedyes- the value is saved and replicated, but nodes keep running with the previous value until thesmrtcservice is restarted on each of them
Durations are written as <integer><unit> (ms, s, m, h, d), and sizes as <integer><unit> with binary multipliers (b, kb, mb, gb, tb). Fractional values are not supported.
Event Ingestion
| Parameter | Description | Default | Restart |
|---|---|---|---|
ingest.queue_size | In-memory ingest queue capacity, in events | 1000000 | yes |
ingest.warmup_mode | Ingestion behavior while the node warms up: block rejects events, accept admits them | block | yes |
ingest.dedupe_ttl | How long a forwarded event identifier is stored for duplicate suppression | 600s | no |
ingest.dedupe_max_entries | Maximum number of identifiers in the duplicate table | 1000000 | yes |
ingest.event_sample_size | Number of event samples stored for each alias; 0 disables sampling, and the maximum is 1000 | 10 | no |
Cluster Coordination
| Parameter | Description | Default | Restart |
|---|---|---|---|
cluster.heartbeat | Interval between node heartbeat messages | 3s | yes |
cluster.peer_timeout | Timeout after which a node is considered unavailable; must exceed cluster.heartbeat | 30s | yes |
cluster.lease | Ownership lease duration used for cluster coordination | 30s | yes |
Data Retention
| Parameter | Description | Default | Restart |
|---|---|---|---|
retention.fires | Firing retention period | 30d | no |
retention.processed_sources | How long processed source statistics survive after the last event | 30d | no |
Both retention periods must be whole multiples of a day: a value such as 36h is rejected.
Negative Filter
| Parameter | Description | Default | Restart |
|---|---|---|---|
negative_filter.enabled | Enables the negative filter | true | yes |
negative_filter.budget | Filter memory budget; at least 1mb | 256mb | yes |
negative_filter.pending_spill | Volume of in-flight pending writes held in memory before they spill to disk; at least 1mb | 32mb | yes |
Time Model
| Parameter | Description | Default | Restart |
|---|---|---|---|
time_model.max_skew_fwd | How far into the future an event timestamp may run before it is clamped to receipt time | 300s | no |
time_model.max_lag_back_floor | Floor on how far back an event timestamp may lag; the effective bound is the greater of the rule window and this value | 3600s | no |
time_model.retention_grace_floor | Minimum grace added to a rule window before its correlation state is reaped | 30s | no |
time_model.retention_grace_ceil | Maximum such grace; must not be below time_model.retention_grace_floor | 300s | no |
Correlation
| Parameter | Description | Default | Restart |
|---|---|---|---|
correlation.max_partials_per_rule | Limit on concurrent incomplete imperative chains per rule | 4096 | no |
correlation.reserved_step_pool | Slots reserved for advancing an existing chain instead of starting a new one | 512 | no |
correlation.global_partial_ceiling | Node-wide limit on incomplete chains across all rules | 1000000 | no |
correlation.max_aggregation_set_per_group | Limit on the distinct values a dc or values metric tracks per group | 100000 | no |
Search Anywhere Framework Connection
| Parameter | Description | Default | Restart |
|---|---|---|---|
smart_monitor.use_tls | Uses TLS when dispatching firings to Search Anywhere Framework | true | no |
smart_monitor.tls_insecure_skip_verify | Skips Search Anywhere Framework certificate verification | false | no |
smart_monitor.dispatch_concurrency | Number of in-flight firing dispatch requests; allowed range is 4 to 1024 | 128 | no |
smart_monitor.tls_insecure_skip_verify disables Search Anywhere Framework certificate authenticity verification.