OCI Monitoring Query Language (MQL) Reference

Last verified: 2026-09-30 against the Monitoring Query Language (MQL) Reference and "Creating an Absence Alarm" on docs.oracle.com.

Shape of a query

MetricName[interval]{dimension filters}.groupingFunction(...).statistic() predicate

Example: CpuUtilization[5m]{availabilityDomain = "VeXI:PHX-AD-1"}.groupBy(resourceId).max() > 85

Interval

1m–60m, 1h–24h, or 1d. For alarms the interval must be at least the metric's emission frequency (hourly metrics need [1h]). Alarm query resolution is always 1 minute.

Dimension filters

Form	Meaning
`{resourceId = "ocid1..."}`	exact match
`{resourceId != "ocid1..."}`	not equal
`{dim1 = "a", dim2 = "b"}`	AND of filters
`{resourceDisplayName =~ "web-*"}`	fuzzy match, `*` wildcard
`{resourceDisplayName =~ "ol8	ol7"}`

=~ exists only in MQL. Do not copy it into IAM policies.

Statistics

absent(), avg(), count(), first(), increment(), last(), max(), mean(), min(), percentile(p) with 0 < p < 1 (e.g. percentile(0.9)), rate() (per-second rate over the interval), sum().

absent() returns 1 when the metric is absent for the whole interval, 0 when present, and stops producing values after the absence detection period (default 2h; absent(20m) … absent(3d)).

Predicates

Operator	Meaning
`>`, `>=`, `==`, `!=`, `<`, `<=`	comparison
`in (a, b)`	between a and b, inclusive
`not in (a, b)`	outside a and b, inclusive

Grouping

groupBy(dim1, dim2) — aggregate per distinct combination of the listed dimensions.
grouping() — aggregate all streams into one.

Arithmetic and joins

+ - * / % between queries or with constants, e.g. CpuUtilization[1m].mean() * 100.
&& and || join whole queries: CpuUtilization[1m]{faultDomain =~ "FAULT-DOMAIN-1|FAULT-DOMAIN-2"}.mean() > 80 || MemoryUtilization[1m]{faultDomain =~ "FAULT-DOMAIN-1|FAULT-DOMAIN-2"}.mean() > 80
They are invalid inside a dimension set.

Verified examples

# Sustained CPU on one instance
CpuUtilization[5m]{resourceId = "<instance-ocid>"}.mean() > 80

# Instances named ol8* or ol7* with low minimum CPU
CpuUtilization[1m]{resourceDisplayName =~ "ol8|ol7"}.min() >= 20

# Host (or agent) stopped reporting for 20 minutes
CpuUtilization[1m]{resourceId = "<instance-ocid>"}.groupBy(resourceId).absent(20m)

# ADB storage (hourly metric)
StorageUtilization[1h]{resourceId = "<adb-ocid>"}.max() > 85

Alarm CLI flags worth knowing (OCI CLI 3.94.1)

--query-text, --namespace, --metric-compartment-id, --pending-duration (ISO-8601, e.g. PT5M), --severity, --destinations, --resolution, --repeat-notification-duration, --is-notifications-per-metric-dimension-enabled, --evaluation-slack-duration, --rule-name.

Official sources

MQL reference: https://docs.oracle.com/en-us/iaas/Content/Monitoring/Reference/mql.htm
Absence alarms: https://docs.oracle.com/en-us/iaas/Content/Monitoring/Tasks/create-alarm-absence.htm
Compute agent metrics: https://docs.oracle.com/en-us/iaas/Content/Compute/References/computemetrics.htm
ADB metrics: https://docs.oracle.com/en-us/iaas/autonomous-database-serverless/doc/autonomous-monitor-metrics-list.html
Load balancer metrics: https://docs.oracle.com/en-us/iaas/Content/Balance/Reference/loadbalancermetrics.htm

/monitoring-operations

referencesoci-monitoring-reference.md